1.\" $OpenBSD: pf.conf.5,v 1.406 2009/01/31 19:37:12 sobrado Exp $ 2.\" 3.\" Copyright (c) 2002, Daniel Hartmeier 4.\" All rights reserved. 5.\" 6.\" Redistribution and use in source and binary forms, with or without 7.\" modification, are permitted provided that the following conditions 8.\" are met: 9.\" 10.\" - Redistributions of source code must retain the above copyright 11.\" notice, this list of conditions and the following disclaimer. 12.\" - Redistributions in binary form must reproduce the above 13.\" copyright notice, this list of conditions and the following 14.\" disclaimer in the documentation and/or other materials provided 15.\" with the distribution. 16.\" 17.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 18.\" "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 19.\" LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 20.\" FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 21.\" COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 22.\" INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 23.\" BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 24.\" LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 25.\" CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 27.\" ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 28.\" POSSIBILITY OF SUCH DAMAGE. 29.\" 30.Dd April 9, 2025 31.Dt PF.CONF 5 32.Os 33.Sh NAME 34.Nm pf.conf 35.Nd packet filter configuration file 36.Sh DESCRIPTION 37The 38.Xr pf 4 39packet filter modifies, drops or passes packets according to rules or 40definitions specified in 41.Nm pf.conf . 42.Sh STATEMENT ORDER 43There are eight types of statements in 44.Nm pf.conf : 45.Bl -tag -width xxxx 46.It Cm Macros 47User-defined variables may be defined and used later, simplifying 48the configuration file. 49Macros must be defined before they are referenced in 50.Nm pf.conf . 51.It Cm Tables 52Tables provide a mechanism for increasing the performance and flexibility of 53rules with large numbers of source or destination addresses. 54.It Cm Options 55Options tune the behaviour of the packet filtering engine. 56.It Cm Ethernet Filtering 57Ethernet filtering provides rule-based blocking or passing of Ethernet packets. 58.It Cm Traffic Normalization Li (e.g. Em scrub ) 59Traffic normalization protects internal machines against inconsistencies 60in Internet protocols and implementations. 61.It Cm Queueing 62Queueing provides rule-based bandwidth control. 63.It Cm Translation Li (Various forms of NAT) 64Translation rules specify how addresses are to be mapped or redirected to 65other addresses. 66.It Cm Packet Filtering 67Packet filtering provides rule-based blocking or passing of packets. 68.El 69.Pp 70With the exception of 71.Cm macros 72and 73.Cm tables , 74the types of statements should be grouped and appear in 75.Nm pf.conf 76in the order shown above, as this matches the operation of the underlying 77packet filtering engine. 78By default 79.Xr pfctl 8 80enforces this order (see 81.Ar set require-order 82below). 83.Pp 84Comments can be put anywhere in the file using a hash mark 85.Pq Sq # , 86and extend to the end of the current line. 87.Pp 88Additional configuration files can be included with the 89.Ic include 90keyword, for example: 91.Bd -literal -offset indent 92include "/etc/pf/sub.filter.conf" 93.Ed 94.Sh MACROS 95Macros can be defined that will later be expanded in context. 96Macro names must start with a letter, and may contain letters, digits 97and underscores. 98Macro names may not be reserved words (for example 99.Ar pass , 100.Ar in , 101.Ar out ) . 102Macros are not expanded inside quotes. 103.Pp 104For example, 105.Bd -literal -offset indent 106ext_if = \&"kue0\&" 107all_ifs = \&"{\&" $ext_if lo0 \&"}\&" 108pass out on $ext_if from any to any 109pass in on $ext_if proto tcp from any to any port 25 110.Ed 111.Sh TABLES 112Tables are named structures which can hold a collection of addresses and 113networks. 114Lookups against tables in 115.Xr pf 4 116are relatively fast, making a single rule with tables much more efficient, 117in terms of 118processor usage and memory consumption, than a large number of rules which 119differ only in IP address (either created explicitly or automatically by rule 120expansion). 121.Pp 122Tables can be used as the source or destination of filter rules, 123.Ar scrub 124rules 125or 126translation rules such as 127.Ar nat 128or 129.Ar rdr 130(see below for details on the various rule types). 131Tables can also be used for the redirect address of 132.Ar nat 133and 134.Ar rdr 135and in the routing options of filter rules, but not for 136.Ar bitmask 137pools. 138.Pp 139Tables can be defined with any of the following 140.Xr pfctl 8 141mechanisms. 142As with macros, reserved words may not be used as table names. 143.Bl -tag -width "manually" 144.It Ar manually 145Persistent tables can be manually created with the 146.Ar add 147or 148.Ar replace 149option of 150.Xr pfctl 8 , 151before or after the ruleset has been loaded. 152.It Pa pf.conf 153Table definitions can be placed directly in this file, and loaded at the 154same time as other rules are loaded, atomically. 155Table definitions inside 156.Nm pf.conf 157use the 158.Ar table 159statement, and are especially useful to define non-persistent tables. 160The contents of a pre-existing table defined without a list of addresses 161to initialize it is not altered when 162.Nm pf.conf 163is loaded. 164A table initialized with the empty list, 165.Li { } , 166will be cleared on load. 167.El 168.Pp 169Tables may be defined with the following attributes: 170.Bl -tag -width counters 171.It Ar persist 172The 173.Ar persist 174flag forces the kernel to keep the table even when no rules refer to it. 175If the flag is not set, the kernel will automatically remove the table 176when the last rule referring to it is flushed. 177.It Ar const 178The 179.Ar const 180flag prevents the user from altering the contents of the table once it 181has been created. 182Without that flag, 183.Xr pfctl 8 184can be used to add or remove addresses from the table at any time, even 185when running with 186.Xr securelevel 7 187= 2. 188.It Ar counters 189The 190.Ar counters 191flag enables per-address packet and byte counters which can be displayed with 192.Xr pfctl 8 . 193Note that this feature carries significant memory overhead for large tables. 194.El 195.Pp 196For example, 197.Bd -literal -offset indent 198table <private> const { 10/8, 172.16/12, 192.168/16 } 199table <badhosts> persist 200block on fxp0 from { <private>, <badhosts> } to any 201.Ed 202.Pp 203creates a table called private, to hold RFC 1918 private network 204blocks, and a table called badhosts, which is initially empty. 205A filter rule is set up to block all traffic coming from addresses listed in 206either table. 207The private table cannot have its contents changed and the badhosts table 208will exist even when no active filter rules reference it. 209Addresses may later be added to the badhosts table, so that traffic from 210these hosts can be blocked by using 211.Bd -literal -offset indent 212# pfctl -t badhosts -Tadd 204.92.77.111 213.Ed 214.Pp 215A table can also be initialized with an address list specified in one or more 216external files, using the following syntax: 217.Bd -literal -offset indent 218table <spam> persist file \&"/etc/spammers\&" file \&"/etc/openrelays\&" 219block on fxp0 from <spam> to any 220.Ed 221.Pp 222The files 223.Pa /etc/spammers 224and 225.Pa /etc/openrelays 226list IP addresses, one per line. 227Any lines beginning with a # are treated as comments and ignored. 228In addition to being specified by IP address, hosts may also be 229specified by their hostname. 230When the resolver is called to add a hostname to a table, 231.Em all 232resulting IPv4 and IPv6 addresses are placed into the table. 233IP addresses can also be entered in a table by specifying a valid interface 234name, a valid interface group or the 235.Em self 236keyword, in which case all addresses assigned to the interface(s) will be 237added to the table. 238.Sh OPTIONS 239.Xr pf 4 240may be tuned for various situations using the 241.Ar set 242command. 243.Bl -tag -width xxxx 244.It Ar set timeout 245.Pp 246.Bl -tag -width "src.track" -compact 247.It Ar interval 248Interval between purging expired states and fragments. 249.It Ar frag 250Seconds before an unassembled fragment is expired. 251.It Ar src.track 252Length of time to retain a source tracking entry after the last state 253expires. 254.El 255.Pp 256When a packet matches a stateful connection, the seconds to live for the 257connection will be updated to that of the 258.Ar proto.modifier 259which corresponds to the connection state. 260Each packet which matches this state will reset the TTL. 261Tuning these values may improve the performance of the 262firewall at the risk of dropping valid idle connections. 263.Pp 264.Bl -tag -width xxxx -compact 265.It Ar tcp.first 266The state after the first packet. 267.It Ar tcp.opening 268The state after the second packet but before both endpoints have 269acknowledged the connection. 270.It Ar tcp.established 271The fully established state. 272.It Ar tcp.closing 273The state after the first FIN has been sent. 274.It Ar tcp.finwait 275The state after both FINs have been exchanged and the connection is closed. 276Some hosts (notably web servers on Solaris) send TCP packets even after closing 277the connection. 278Increasing 279.Ar tcp.finwait 280(and possibly 281.Ar tcp.closing ) 282can prevent blocking of such packets. 283.It Ar tcp.closed 284The state after one endpoint sends an RST. 285.El 286.Pp 287SCTP timeout are handled similar to TCP, but with its own set of states: 288.Pp 289.Bl -tag -width xxxx -compact 290.It Ar sctp.first 291The state after the first packet. 292.It Ar sctp.opening 293The state before the destination host ever sends a packet. 294.It Ar sctp.established 295The fully established state. 296.It Ar sctp.closing 297The state after the first SHUTDOWN chunk has been sent. 298.It Ar sctp.closed 299The state after SHUTDOWN_ACK has been exchanged and the connection is closed. 300.El 301.Pp 302ICMP and UDP are handled in a fashion similar to TCP, but with a much more 303limited set of states: 304.Pp 305.Bl -tag -width xxxx -compact 306.It Ar udp.first 307The state after the first packet. 308.It Ar udp.single 309The state if the source host sends more than one packet but the destination 310host has never sent one back. 311.It Ar udp.multiple 312The state if both hosts have sent packets. 313.It Ar icmp.first 314The state after the first packet. 315.It Ar icmp.error 316The state after an ICMP error came back in response to an ICMP packet. 317.El 318.Pp 319Other protocols are handled similarly to UDP: 320.Pp 321.Bl -tag -width xxxx -compact 322.It Ar other.first 323.It Ar other.single 324.It Ar other.multiple 325.El 326.Pp 327Timeout values can be reduced adaptively as the number of state table 328entries grows. 329.Pp 330.Bl -tag -width xxxx -compact 331.It Ar adaptive.start 332When the number of state entries exceeds this value, adaptive scaling 333begins. 334All timeout values are scaled linearly with factor 335(adaptive.end - number of states) / (adaptive.end - adaptive.start). 336.It Ar adaptive.end 337When reaching this number of state entries, all timeout values become 338zero, effectively purging all state entries immediately. 339This value is used to define the scale factor, it should not actually 340be reached (set a lower state limit, see below). 341.El 342.Pp 343Adaptive timeouts are enabled by default, with an adaptive.start value 344equal to 60% of the state limit, and an adaptive.end value equal to 345120% of the state limit. 346They can be disabled by setting both adaptive.start and adaptive.end to 0. 347.Pp 348The adaptive timeout values can be defined both globally and for each rule. 349When used on a per-rule basis, the values relate to the number of 350states created by the rule, otherwise to the total number of 351states. 352.Pp 353For example: 354.Bd -literal -offset indent 355set timeout tcp.first 120 356set timeout tcp.established 86400 357set timeout { adaptive.start 6000, adaptive.end 12000 } 358set limit states 10000 359.Ed 360.Pp 361With 9000 state table entries, the timeout values are scaled to 50% 362(tcp.first 60, tcp.established 43200). 363.It Ar set loginterface 364Enable collection of packet and byte count statistics for the given 365interface or interface group. 366These statistics can be viewed using 367.Bd -literal -offset indent 368# pfctl -s info 369.Ed 370.Pp 371In this example 372.Xr pf 4 373collects statistics on the interface named dc0: 374.Bd -literal -offset indent 375set loginterface dc0 376.Ed 377.Pp 378One can disable the loginterface using: 379.Bd -literal -offset indent 380set loginterface none 381.Ed 382.It Ar set limit 383Sets hard limits on the memory pools used by the packet filter. 384See 385.Xr zone 9 386for an explanation of memory pools. 387.Pp 388For example, 389.Bd -literal -offset indent 390set limit states 20000 391.Ed 392.Pp 393sets the maximum number of entries in the memory pool used by state table 394entries (generated by 395.Ar pass 396rules which do not specify 397.Ar no state ) 398to 20000. 399Using 400.Bd -literal -offset indent 401set limit frags 20000 402.Ed 403.Pp 404sets the maximum number of entries in the memory pool used for fragment 405reassembly (generated by the 406.Ar set reassemble 407option or 408.Ar scrub 409rules) to 20000. 410Using 411.Bd -literal -offset indent 412set limit src-nodes 2000 413.Ed 414.Pp 415sets the maximum number of entries in the memory pool used for tracking 416source IP addresses (generated by the 417.Ar sticky-address 418and 419.Ar src.track 420options) to 2000. 421Using 422.Bd -literal -offset indent 423set limit table-entries 100000 424.Ed 425.Pp 426sets the limit on the overall number of addresses that can be stored 427in tables to 100000. 428.Pp 429Various limits can be combined on a single line: 430.Bd -literal -offset indent 431set limit { states 20000, frags 20000, src-nodes 2000 } 432.Ed 433.It Ar set ruleset-optimization 434.Bl -tag -width xxxxxxxx -compact 435.It Ar none 436Disable the ruleset optimizer. 437.It Ar basic 438Enable basic ruleset optimization. 439This is the default behaviour. 440Basic ruleset optimization does four things to improve the 441performance of ruleset evaluations: 442.Pp 443.Bl -enum -compact 444.It 445remove duplicate rules 446.It 447remove rules that are a subset of another rule 448.It 449combine multiple rules into a table when advantageous 450.It 451re-order the rules to improve evaluation performance 452.El 453.Pp 454.It Ar profile 455Uses the currently loaded ruleset as a feedback profile to tailor the 456ordering of quick rules to actual network traffic. 457.El 458.Pp 459It is important to note that the ruleset optimizer will modify the ruleset 460to improve performance. 461A side effect of the ruleset modification is that per-rule accounting 462statistics will have different meanings than before. 463If per-rule accounting is important for billing purposes or whatnot, 464either the ruleset optimizer should not be used or a label field should 465be added to all of the accounting rules to act as optimization barriers. 466.Pp 467Optimization can also be set as a command-line argument to 468.Xr pfctl 8 , 469overriding the settings in 470.Nm . 471.It Ar set optimization 472Optimize state timeouts for one of the following network environments: 473.Pp 474.Bl -tag -width xxxx -compact 475.It Ar normal 476A normal network environment. 477Suitable for almost all networks. 478.It Ar high-latency 479A high-latency environment (such as a satellite connection). 480.It Ar satellite 481Alias for 482.Ar high-latency . 483.It Ar aggressive 484Aggressively expire connections. 485This can greatly reduce the memory usage of the firewall at the cost of 486dropping idle connections early. 487.It Ar conservative 488Extremely conservative settings. 489Avoid dropping legitimate connections at the 490expense of greater memory utilization (possibly much greater on a busy 491network) and slightly increased processor utilization. 492.El 493.Pp 494For example: 495.Bd -literal -offset indent 496set optimization aggressive 497.Ed 498.It Ar set reassemble yes | no Op Cm no-df 499The 500.Cm reassemble 501option is used to enable or disable the reassembly of fragmented packets, 502and can be set to 503.Cm yes 504or 505.Cm no . 506If 507.Cm no-df 508is also specified, fragments with the 509.Dq dont-fragment 510bit set are reassembled too, 511instead of being dropped; 512the reassembled packet will have the 513.Dq dont-fragment 514bit cleared. 515The default value is 516.Cm no . 517.Pp 518This option is ignored if there are pre-FreeBSD 14 519.Cm scrub 520rules present. 521.It Ar set block-policy 522The 523.Ar block-policy 524option sets the default behaviour for the packet 525.Ar block 526action: 527.Pp 528.Bl -tag -width xxxxxxxx -compact 529.It Ar drop 530Packet is silently dropped. 531.It Ar return 532A TCP RST is returned for blocked TCP packets, 533an SCTP ABORT chunk is returned for blocked SCTP packets, 534an ICMP UNREACHABLE is returned for blocked UDP packets, 535and all other packets are silently dropped. 536.El 537.Pp 538For example: 539.Bd -literal -offset indent 540set block-policy return 541.Ed 542.It Ar set fail-policy 543The 544.Ar fail-policy 545option sets the behaviour of rules which should pass a packet but were 546unable to do so. 547This might happen when a nat or route-to rule uses an empty table as list 548of targets or if a rule fails to create state or source node. 549The following 550.Ar block 551actions are possible: 552.Pp 553.Bl -tag -width xxxxxxxx -compact 554.It Ar drop 555Incoming packet is silently dropped. 556.It Ar return 557Incoming packet is dropped and TCP RST is returned for TCP packets, 558an SCTP ABORT chunk is returned for blocked SCTP packets, 559an ICMP UNREACHABLE is returned for UDP packets, 560and no response is sent for other packets. 561.El 562.Pp 563For example: 564.Bd -literal -offset indent 565set fail-policy return 566.Ed 567.It Ar set state-policy 568The 569.Ar state-policy 570option sets the default behaviour for states: 571.Pp 572.Bl -tag -width group-bound -compact 573.It Ar if-bound 574States are bound to interface. 575.It Ar floating 576States can match packets on any interfaces (the default). 577.El 578.Pp 579For example: 580.Bd -literal -offset indent 581set state-policy if-bound 582.Ed 583.It Ar set syncookies never | always | adaptive 584When 585.Cm syncookies 586are active, pf will answer each incoming TCP SYN with a syncookie SYNACK, 587without allocating any resources. 588Upon reception of the client's ACK in response to the syncookie 589SYNACK, pf will evaluate the ruleset and create state if the ruleset 590permits it, complete the three way handshake with the target host and 591continue the connection with synproxy in place. 592This allows pf to be resilient against large synflood attacks which would 593run the state table against its limits otherwise. 594Due to the blind answers to every incoming SYN syncookies share the caveats of 595synproxy, namely seemingly accepting connections that will be dropped later on. 596.Pp 597.Bl -tag -width adaptive -compact 598.It Cm never 599pf will never send syncookie SYNACKs (the default). 600.It Cm always 601pf will always send syncookie SYNACKs. 602.It Cm adaptive 603pf will enable syncookie mode when a given percentage of the state table 604is used up by half-open TCP connections, as in, those that saw the initial 605SYN but didn't finish the three way handshake. 606The thresholds for entering and leaving syncookie mode can be specified using 607.Bd -literal -offset indent 608set syncookies adaptive (start 25%, end 12%) 609.Ed 610.El 611.It Ar set state-defaults 612The 613.Ar state-defaults 614option sets the state options for states created from rules 615without an explicit 616.Ar keep state . 617For example: 618.Bd -literal -offset indent 619set state-defaults no-sync 620.Ed 621.It Ar set hostid 622The 32-bit 623.Ar hostid 624identifies this firewall's state table entries to other firewalls 625in a 626.Xr pfsync 4 627failover cluster. 628By default the hostid is set to a pseudo-random value, however it may be 629desirable to manually configure it, for example to more easily identify the 630source of state table entries. 631.Bd -literal -offset indent 632set hostid 1 633.Ed 634.Pp 635The hostid may be specified in either decimal or hexadecimal. 636.It Ar set require-order 637By default 638.Xr pfctl 8 639enforces an ordering of the statement types in the ruleset to: 640.Em options , 641.Em normalization , 642.Em queueing , 643.Em translation , 644.Em filtering . 645Setting this option to 646.Ar no 647disables this enforcement. 648There may be non-trivial and non-obvious implications to an out of 649order ruleset. 650Consider carefully before disabling the order enforcement. 651.It Ar set fingerprints 652Load fingerprints of known operating systems from the given filename. 653By default fingerprints of known operating systems are automatically 654loaded from 655.Xr pf.os 5 656in 657.Pa /etc 658but can be overridden via this option. 659Setting this option may leave a small period of time where the fingerprints 660referenced by the currently active ruleset are inconsistent until the new 661ruleset finishes loading. 662.Pp 663For example: 664.Pp 665.Dl set fingerprints \&"/etc/pf.os.devel\&" 666.It Ar set skip on Aq Ar ifspec 667List interfaces for which packets should not be filtered. 668Packets passing in or out on such interfaces are passed as if pf was 669disabled, i.e. pf does not process them in any way. 670This can be useful on loopback and other virtual interfaces, when 671packet filtering is not desired and can have unexpected effects. 672For example: 673.Pp 674.Dl set skip on lo0 675.It Ar set debug 676Set the debug 677.Ar level 678to one of the following: 679.Pp 680.Bl -tag -width xxxxxxxxxxxx -compact 681.It Ar none 682Don't generate debug messages. 683.It Ar urgent 684Generate debug messages only for serious errors. 685.It Ar misc 686Generate debug messages for various errors. 687.It Ar loud 688Generate debug messages for common conditions. 689.El 690.It Ar set keepcounters 691Preserve rule counters across rule updates. 692Usually rule counters are reset to zero on every update of the ruleset. 693With 694.Ar keepcounters 695set pf will attempt to find matching rules between old and new rulesets 696and preserve the rule counters. 697.El 698.Sh ETHERNET FILTERING 699.Xr pf 4 700has the ability to 701.Ar block 702and 703.Ar pass 704packets based on attributes of their Ethernet (layer 2) header. 705.Pp 706For each packet processed by the packet filter, the filter rules are 707evaluated in sequential order, from first to last. 708The last matching rule decides what action is taken. 709If no rule matches the packet, the default action is to pass 710the packet without creating a state. 711.Pp 712The following actions can be used in the filter: 713.Bl -tag -width xxxx 714.It Ar block 715The packet is blocked. 716Unlike for layer 3 traffic the packet is always silently dropped. 717.It Ar pass 718The packet is passed; 719no state is created for layer 2 traffic. 720.El 721.Sh PARAMETERS 722The rule parameters specify the packets to which a rule applies. 723A packet always comes in on, or goes out through, one interface. 724Most parameters are optional. 725If a parameter is specified, the rule only applies to packets with 726matching attributes. 727Certain parameters can be expressed as lists, in which case 728.Xr pfctl 8 729generates all needed rule combinations. 730.Bl -tag -width xxxx 731.It Ar in No or Ar out 732This rule applies to incoming or outgoing packets. 733If neither 734.Ar in 735nor 736.Ar out 737are specified, the rule will match packets in both directions. 738.It Ar quick 739If a packet matches a rule which has the 740.Ar quick 741option set, this rule 742is considered the last matching rule, and evaluation of subsequent rules 743is skipped. 744.It Ar on Aq Ar ifspec 745This rule applies only to packets coming in on, or going out through, this 746particular interface or interface group. 747For more information on interface groups, 748see the 749.Ic group 750keyword in 751.Xr ifconfig 8 . 752.Ar any 753will match any existing interface except loopback ones. 754.It Ar bridge-to Aq interface 755Packets matching this rule will be sent out of the specified interface without 756further processing. 757.It Ar proto Aq Ar protocol 758This rule applies only to packets of this protocol. 759Note that Ethernet protocol numbers are different from those used in 760.Xr ip 4 761and 762.Xr ip6 4 . 763.It Xo 764.Ar from Aq Ar source 765.Ar to Aq Ar dest 766.Xc 767This rule applies only to packets with the specified source and destination 768MAC addresses. 769.It Xo Ar queue Aq Ar queue 770.Xc 771Packets matching this rule will be assigned to the specified queue. 772See 773.Sx QUEUEING 774for setup details. 775.Pp 776.It Ar tag Aq Ar string 777Packets matching this rule will be tagged with the 778specified string. 779The tag acts as an internal marker that can be used to 780identify these packets later on. 781This can be used, for example, to provide trust between 782interfaces and to determine if packets have been 783processed by translation rules. 784Tags are 785.Qq sticky , 786meaning that the packet will be tagged even if the rule 787is not the last matching rule. 788Further matching rules can replace the tag with a 789new one but will not remove a previously applied tag. 790A packet is only ever assigned one tag at a time. 791.It Ar tagged Aq Ar string 792Used to specify that packets must already be tagged with the given tag in order 793to match the rule. 794Inverse tag matching can also be done by specifying the ! operator before the 795tagged keyword. 796.El 797.Sh TRAFFIC NORMALIZATION 798Traffic normalization is a broad umbrella term 799for aspects of the packet filter which deal with 800verifying packets, packet fragments, spoofed traffic, 801and other irregularities. 802.Ss Scrub 803Scrub involves sanitising packet content in such a way 804that there are no ambiguities in packet interpretation on the receiving side. 805It is invoked with the 806.Cm scrub 807option, added to filter rules. 808.Pp 809Parameters are specified enclosed in parentheses. 810At least one of the following parameters must be specified: 811.Bl -tag -width xxxx 812.It Ar no-df 813Clears the 814.Ar dont-fragment 815bit from a matching IP packet. 816Some operating systems are known to generate fragmented packets with the 817.Ar dont-fragment 818bit set. 819This is particularly true with NFS. 820.Ar Scrub 821will drop such fragmented 822.Ar dont-fragment 823packets unless 824.Ar no-df 825is specified. 826.Pp 827Unfortunately some operating systems also generate their 828.Ar dont-fragment 829packets with a zero IP identification field. 830Clearing the 831.Ar dont-fragment 832bit on packets with a zero IP ID may cause deleterious results if an 833upstream router later fragments the packet. 834Using the 835.Ar random-id 836modifier (see below) is recommended in combination with the 837.Ar no-df 838modifier to ensure unique IP identifiers. 839.It Ar min-ttl Aq Ar number 840Enforces a minimum TTL for matching IP packets. 841.It Ar max-mss Aq Ar number 842Enforces a maximum MSS for matching TCP packets. 843.It Xo Ar set-tos Aq Ar string 844.No \*(Ba Aq Ar number 845.Xc 846Enforces a 847.Em TOS 848for matching IP packets. 849.Em TOS 850may be 851given as one of 852.Ar critical , 853.Ar inetcontrol , 854.Ar lowdelay , 855.Ar netcontrol , 856.Ar throughput , 857.Ar reliability , 858or one of the DiffServ Code Points: 859.Ar ef , 860.Ar va , 861.Ar af11 No ... Ar af43 , 862.Ar cs0 No ... Ar cs7 ; 863or as either hex or decimal. 864.It Ar random-id 865Replaces the IP identification field with random values to compensate 866for predictable values generated by many hosts. 867This option only applies to packets that are not fragmented 868after the optional fragment reassembly. 869.It Ar reassemble tcp 870Statefully normalizes TCP connections. 871.Ar reassemble tcp 872performs the following normalizations: 873.Pp 874.Bl -tag -width timeout -compact 875.It ttl 876Neither side of the connection is allowed to reduce their IP TTL. 877An attacker may send a packet such that it reaches the firewall, affects 878the firewall state, and expires before reaching the destination host. 879.Ar reassemble tcp 880will raise the TTL of all packets back up to the highest value seen on 881the connection. 882.It timestamp modulation 883Modern TCP stacks will send a timestamp on every TCP packet and echo 884the other endpoint's timestamp back to them. 885Many operating systems will merely start the timestamp at zero when 886first booted, and increment it several times a second. 887The uptime of the host can be deduced by reading the timestamp and multiplying 888by a constant. 889Also observing several different timestamps can be used to count hosts 890behind a NAT device. 891And spoofing TCP packets into a connection requires knowing or guessing 892valid timestamps. 893Timestamps merely need to be monotonically increasing and not derived off a 894guessable base time. 895.Ar reassemble tcp 896will cause 897.Ar scrub 898to modulate the TCP timestamps with a random number. 899.It extended PAWS checks 900There is a problem with TCP on long fat pipes, in that a packet might get 901delayed for longer than it takes the connection to wrap its 32-bit sequence 902space. 903In such an occurrence, the old packet would be indistinguishable from a 904new packet and would be accepted as such. 905The solution to this is called PAWS: Protection Against Wrapped Sequence 906numbers. 907It protects against it by making sure the timestamp on each packet does 908not go backwards. 909.Ar reassemble tcp 910also makes sure the timestamp on the packet does not go forward more 911than the RFC allows. 912By doing this, 913.Xr pf 4 914artificially extends the security of TCP sequence numbers by 10 to 18 915bits when the host uses appropriately randomized timestamps, since a 916blind attacker would have to guess the timestamp as well. 917.El 918.El 919.Pp 920For example, 921.Bd -literal -offset indent 922match in all scrub (no-df random-id max-mss 1440) 923.Ed 924.Ss Scrub ruleset (pre-FreeBSD 14) 925In order to maintain compatibility with older releases of FreeBSD 926.Ar scrub 927rules can also be specified in their own ruleset. 928In such case they are invoked with the 929.Ar scrub 930directive. 931If there are such rules present they determine packet reassembly behaviour. 932When no such rules are present the option 933.Ar set reassembly 934takes precedence. 935The 936.Ar scrub 937rules can take all parameters specified above for a 938.Ar scrub 939option of filter rules and 2 more parameters controlling fragment reassembly: 940.Bl -tag -width xxxx 941.It Ar fragment reassemble 942Using 943.Ar scrub 944rules, fragments can be reassembled by normalization. 945In this case, fragments are buffered until they form a complete 946packet, and only the completed packet is passed on to the filter. 947The advantage is that filter rules have to deal only with complete 948packets, and can ignore fragments. 949The drawback of caching fragments is the additional memory cost. 950This is the default behaviour unless no fragment reassemble is specified. 951.It Ar no fragment reassemble 952Do not reassemble fragments. 953.El 954.Pp 955For example, 956.Bd -literal -offset indent 957scrub in on $ext_if all fragment reassemble 958.Ed 959.Pp 960The 961.Ar no 962option prefixed to a scrub rule causes matching packets to remain unscrubbed, 963much in the same way as 964.Ar drop quick 965works in the packet filter (see below). 966This mechanism should be used when it is necessary to exclude specific packets 967from broader scrub rules. 968.Pp 969.Ar scrub 970rules in the 971.Ar scrub 972ruleset are evaluated for every packet before stateful filtering. 973This means excessive usage of them will cause performance penalty. 974.Ar scrub reassemble tcp 975rules must not have the direction (in/out) specified. 976.Sh QUEUEING with ALTQ 977The ALTQ system is currently not available in the GENERIC kernel nor as 978loadable modules. 979In order to use the herein after called queueing options one has to use a 980custom built kernel. 981Please refer to 982.Xr altq 4 983to learn about the related kernel options. 984.Pp 985Packets can be assigned to queues for the purpose of bandwidth 986control. 987At least two declarations are required to configure queues, and later 988any packet filtering rule can reference the defined queues by name. 989During the filtering component of 990.Nm pf.conf , 991the last referenced 992.Ar queue 993name is where any packets from 994.Ar pass 995rules will be queued, while for 996.Ar block 997rules it specifies where any resulting ICMP or TCP RST 998packets should be queued. 999The 1000.Ar scheduler 1001defines the algorithm used to decide which packets get delayed, dropped, or 1002sent out immediately. 1003There are three 1004.Ar schedulers 1005currently supported. 1006.Bl -tag -width xxxx 1007.It Ar cbq 1008Class Based Queueing. 1009.Ar Queues 1010attached to an interface build a tree, thus each 1011.Ar queue 1012can have further child 1013.Ar queues . 1014Each queue can have a 1015.Ar priority 1016and a 1017.Ar bandwidth 1018assigned. 1019.Ar Priority 1020mainly controls the time packets take to get sent out, while 1021.Ar bandwidth 1022has primarily effects on throughput. 1023.Ar cbq 1024achieves both partitioning and sharing of link bandwidth 1025by hierarchically structured classes. 1026Each class has its own 1027.Ar queue 1028and is assigned its share of 1029.Ar bandwidth . 1030A child class can borrow bandwidth from its parent class 1031as long as excess bandwidth is available 1032(see the option 1033.Ar borrow , 1034below). 1035.It Ar priq 1036Priority Queueing. 1037.Ar Queues 1038are flat attached to the interface, thus, 1039.Ar queues 1040cannot have further child 1041.Ar queues . 1042Each 1043.Ar queue 1044has a unique 1045.Ar priority 1046assigned, ranging from 0 to 15. 1047Packets in the 1048.Ar queue 1049with the highest 1050.Ar priority 1051are processed first. 1052.It Ar hfsc 1053Hierarchical Fair Service Curve. 1054.Ar Queues 1055attached to an interface build a tree, thus each 1056.Ar queue 1057can have further child 1058.Ar queues . 1059Each queue can have a 1060.Ar priority 1061and a 1062.Ar bandwidth 1063assigned. 1064.Ar Priority 1065mainly controls the time packets take to get sent out, while 1066.Ar bandwidth 1067primarily affects throughput. 1068.Ar hfsc 1069supports both link-sharing and guaranteed real-time services. 1070It employs a service curve based QoS model, 1071and its unique feature is an ability to decouple 1072.Ar delay 1073and 1074.Ar bandwidth 1075allocation. 1076.El 1077.Pp 1078The interfaces on which queueing should be activated are declared using 1079the 1080.Ar altq on 1081declaration. 1082.Ar altq on 1083has the following keywords: 1084.Bl -tag -width xxxx 1085.It Aq Ar interface 1086Queueing is enabled on the named interface. 1087.It Aq Ar scheduler 1088Specifies which queueing scheduler to use. 1089Currently supported values 1090are 1091.Ar cbq 1092for Class Based Queueing, 1093.Ar priq 1094for Priority Queueing and 1095.Ar hfsc 1096for the Hierarchical Fair Service Curve scheduler. 1097.It Ar bandwidth Aq Ar bw 1098The maximum bitrate for all queues on an 1099interface may be specified using the 1100.Ar bandwidth 1101keyword. 1102The value can be specified as an absolute value or as a 1103percentage of the interface bandwidth. 1104When using an absolute value, the suffixes 1105.Ar b , 1106.Ar Kb , 1107.Ar Mb , 1108and 1109.Ar Gb 1110are used to represent bits, kilobits, megabits, and 1111gigabits per second, respectively. 1112The value must not exceed the interface bandwidth. 1113If 1114.Ar bandwidth 1115is not specified, the interface bandwidth is used 1116(but take note that some interfaces do not know their bandwidth, 1117or can adapt their bandwidth rates). 1118.It Ar qlimit Aq Ar limit 1119The maximum number of packets held in the queue. 1120The default is 50. 1121.It Ar tbrsize Aq Ar size 1122Adjusts the size, in bytes, of the token bucket regulator. 1123If not specified, heuristics based on the 1124interface bandwidth are used to determine the size. 1125.It Ar queue Aq Ar list 1126Defines a list of subqueues to create on an interface. 1127.El 1128.Pp 1129In the following example, the interface dc0 1130should queue up to 5Mbps in four second-level queues using 1131Class Based Queueing. 1132Those four queues will be shown in a later example. 1133.Bd -literal -offset indent 1134altq on dc0 cbq bandwidth 5Mb queue { std, http, mail, ssh } 1135.Ed 1136.Pp 1137Once interfaces are activated for queueing using the 1138.Ar altq 1139directive, a sequence of 1140.Ar queue 1141directives may be defined. 1142The name associated with a 1143.Ar queue 1144must match a queue defined in the 1145.Ar altq 1146directive (e.g. mail), or, except for the 1147.Ar priq 1148.Ar scheduler , 1149in a parent 1150.Ar queue 1151declaration. 1152The following keywords can be used: 1153.Bl -tag -width xxxx 1154.It Ar on Aq Ar interface 1155Specifies the interface the queue operates on. 1156If not given, it operates on all matching interfaces. 1157.It Ar bandwidth Aq Ar bw 1158Specifies the maximum bitrate to be processed by the queue. 1159This value must not exceed the value of the parent 1160.Ar queue 1161and can be specified as an absolute value or a percentage of the parent 1162queue's bandwidth. 1163If not specified, defaults to 100% of the parent queue's bandwidth. 1164The 1165.Ar priq 1166scheduler does not support bandwidth specification. 1167.It Ar priority Aq Ar level 1168Between queues a priority level can be set. 1169For 1170.Ar cbq 1171and 1172.Ar hfsc , 1173the range is 0 to 7 and for 1174.Ar priq , 1175the range is 0 to 15. 1176The default for all is 1. 1177.Ar Priq 1178queues with a higher priority are always served first. 1179.Ar Cbq 1180and 1181.Ar Hfsc 1182queues with a higher priority are preferred in the case of overload. 1183.It Ar qlimit Aq Ar limit 1184The maximum number of packets held in the queue. 1185The default is 50. 1186.El 1187.Pp 1188The 1189.Ar scheduler 1190can get additional parameters with 1191.Xo Aq Ar scheduler 1192.Pf ( Aq Ar parameters ) . 1193.Xc 1194Parameters are as follows: 1195.Bl -tag -width Fl 1196.It Ar default 1197Packets not matched by another queue are assigned to this one. 1198Exactly one default queue is required. 1199.It Ar red 1200Enable RED (Random Early Detection) on this queue. 1201RED drops packets with a probability proportional to the average 1202queue length. 1203.It Ar rio 1204Enables RIO on this queue. 1205RIO is RED with IN/OUT, thus running 1206RED two times more than RIO would achieve the same effect. 1207RIO is currently not supported in the GENERIC kernel. 1208.It Ar ecn 1209Enables ECN (Explicit Congestion Notification) on this queue. 1210ECN implies RED. 1211.El 1212.Pp 1213The 1214.Ar cbq 1215.Ar scheduler 1216supports an additional option: 1217.Bl -tag -width Fl 1218.It Ar borrow 1219The queue can borrow bandwidth from the parent. 1220.El 1221.Pp 1222The 1223.Ar hfsc 1224.Ar scheduler 1225supports some additional options: 1226.Bl -tag -width Fl 1227.It Ar realtime Aq Ar sc 1228The minimum required bandwidth for the queue. 1229.It Ar upperlimit Aq Ar sc 1230The maximum allowed bandwidth for the queue. 1231.It Ar linkshare Aq Ar sc 1232The bandwidth share of a backlogged queue. 1233.El 1234.Pp 1235.Aq Ar sc 1236is an acronym for 1237.Ar service curve . 1238.Pp 1239The format for service curve specifications is 1240.Ar ( m1 , d , m2 ) . 1241.Ar m2 1242controls the bandwidth assigned to the queue. 1243.Ar m1 1244and 1245.Ar d 1246are optional and can be used to control the initial bandwidth assignment. 1247For the first 1248.Ar d 1249milliseconds the queue gets the bandwidth given as 1250.Ar m1 , 1251afterwards the value given in 1252.Ar m2 . 1253.Pp 1254Furthermore, with 1255.Ar cbq 1256and 1257.Ar hfsc , 1258child queues can be specified as in an 1259.Ar altq 1260declaration, thus building a tree of queues using a part of 1261their parent's bandwidth. 1262.Pp 1263Packets can be assigned to queues based on filter rules by using the 1264.Ar queue 1265keyword. 1266Normally only one 1267.Ar queue 1268is specified; when a second one is specified it will instead be used for 1269packets which have a 1270.Em TOS 1271of 1272.Em lowdelay 1273and for TCP ACKs with no data payload. 1274.Pp 1275To continue the previous example, the examples below would specify the 1276four referenced 1277queues, plus a few child queues. 1278Interactive 1279.Xr ssh 1 1280sessions get priority over bulk transfers like 1281.Xr scp 1 1282and 1283.Xr sftp 1 . 1284The queues may then be referenced by filtering rules (see 1285.Sx PACKET FILTERING 1286below). 1287.Bd -literal 1288queue std bandwidth 10% cbq(default) 1289queue http bandwidth 60% priority 2 cbq(borrow red) \e 1290 { employees, developers } 1291queue developers bandwidth 75% cbq(borrow) 1292queue employees bandwidth 15% 1293queue mail bandwidth 10% priority 0 cbq(borrow ecn) 1294queue ssh bandwidth 20% cbq(borrow) { ssh_interactive, ssh_bulk } 1295queue ssh_interactive bandwidth 50% priority 7 cbq(borrow) 1296queue ssh_bulk bandwidth 50% priority 0 cbq(borrow) 1297 1298block return out on dc0 inet all queue std 1299pass out on dc0 inet proto tcp from $developerhosts to any port 80 \e 1300 queue developers 1301pass out on dc0 inet proto tcp from $employeehosts to any port 80 \e 1302 queue employees 1303pass out on dc0 inet proto tcp from any to any port 22 \e 1304 queue(ssh_bulk, ssh_interactive) 1305pass out on dc0 inet proto tcp from any to any port 25 \e 1306 queue mail 1307.Ed 1308.Sh QUEUEING with dummynet 1309Queueing can also be done with 1310.Xr dummynet 4 . 1311Queues and pipes can be created with 1312.Xr dnctl 8 . 1313.Pp 1314Packets can be assigned to queues and pipes using 1315.Ar dnqueue 1316and 1317.Ar dnpipe 1318respectively. 1319.Pp 1320Both 1321.Ar dnqueue 1322and 1323.Ar dnpipe 1324take either a single pipe or queue number or two numbers as arguments. 1325The first pipe or queue number will be used to shape the traffic in the rule 1326direction, the second will be used to shape the traffic in the reverse 1327direction. 1328If the rule does not specify a direction the first packet to create state will 1329be shaped according to the first number, and the response traffic according to 1330the second. 1331.Pp 1332If the 1333.Xr dummynet 4 1334module is not loaded any traffic sent into a queue or pipe will be dropped. 1335.Sh TRANSLATION 1336Translation rules modify either the source or destination address of the 1337packets associated with a stateful connection. 1338A stateful connection is automatically created to track packets matching 1339such a rule as long as they are not blocked by the filtering section of 1340.Nm pf.conf . 1341The translation engine modifies the specified address and/or port in the 1342packet, recalculates IP, TCP and UDP checksums as necessary, and passes 1343it to the packet filter for evaluation. 1344.Pp 1345Since translation occurs before filtering the filter 1346engine will see packets as they look after any 1347addresses and ports have been translated. 1348Filter rules will therefore have to filter based on the translated 1349address and port number. 1350Packets that match a translation rule are only automatically passed if 1351the 1352.Ar pass 1353modifier is given, otherwise they are 1354still subject to 1355.Ar block 1356and 1357.Ar pass 1358rules. 1359.Pp 1360The state entry created permits 1361.Xr pf 4 1362to keep track of the original address for traffic associated with that state 1363and correctly direct return traffic for that connection. 1364.Pp 1365Various types of translation are possible with pf: 1366.Bl -tag -width xxxx 1367.It Ar af-to 1368Translation between different address families (NAT64) is handled 1369using 1370.Ar af-to 1371rules. 1372Because address family translation overrides the routing table, it's 1373only possible to use 1374.Ar af-to 1375on inbound rules, and a source address of the resulting translation 1376must always be specified. 1377.Pp 1378The optional second argument is the host or subnet the original 1379addresses are translated into for the destination. 1380The lowest bits of the original destination address form the host 1381part of the new destination address according to the specified subnet. 1382It is possible to embed a complete IPv4 address into an IPv6 address 1383using a network prefix of /96 or smaller. 1384.Pp 1385When a destination address is not specified it is assumed that the host 1386part is 32-bit long. 1387For IPv6 to IPv4 translation this would mean using only the lower 32 1388bits of the original IPv6 destination address. 1389For IPv4 to IPv6 translation the destination subnet defaults to the 1390subnet of the new IPv6 source address with a prefix length of /96. 1391See RFC 6052 Section 2.2 for details on how the prefix determines the 1392destination address encoding. 1393.Pp 1394For example, the following rules are identical: 1395.Bd -literal -offset indent 1396pass in inet af-to inet6 from 2001:db8::1 to 2001:db8::/96 1397pass in inet af-to inet6 from 2001:db8::1 1398.Ed 1399.Pp 1400In the above example the matching IPv4 packets will be modified to 1401have a source address of 2001:db8::1 and a destination address will 1402get prefixed with 2001:db8::/96, e.g. 198.51.100.100 will be 1403translated to 2001:db8::c633:6464. 1404.Pp 1405In the reverse case the following rules are identical: 1406.Bd -literal -offset indent 1407pass in inet6 af-to inet from 198.51.100.1 to 0.0.0.0/0 1408pass in inet6 af-to inet from 198.51.100.1 1409.Ed 1410.Pp 1411The destination IPv4 address is assumed to be embedded inside the 1412original IPv6 destination address, e.g. 64:ff9b::c633:6464 will be 1413translated to 198.51.100.100. 1414.Pp 1415The current implementation will only extract IPv4 addresses from the 1416IPv6 addresses with a prefix length of /96 and greater. 1417.It Ar binat 1418A 1419.Ar binat 1420rule specifies a bidirectional mapping between an external IP netblock 1421and an internal IP netblock. 1422.It Ar nat 1423A 1424.Ar nat 1425rule specifies that IP addresses are to be changed as the packet 1426traverses the given interface. 1427This technique allows one or more IP addresses 1428on the translating host to support network traffic for a larger range of 1429machines on an "inside" network. 1430Although in theory any IP address can be used on the inside, it is strongly 1431recommended that one of the address ranges defined by RFC 1918 be used. 1432These netblocks are: 1433.Bd -literal 143410.0.0.0 - 10.255.255.255 (all of net 10, i.e., 10/8) 1435172.16.0.0 - 172.31.255.255 (i.e., 172.16/12) 1436192.168.0.0 - 192.168.255.255 (i.e., 192.168/16) 1437.Ed 1438.It Pa rdr 1439The packet is redirected to another destination and possibly a 1440different port. 1441.Ar rdr 1442rules can optionally specify port ranges instead of single ports. 1443rdr ... port 2000:2999 -> ... port 4000 1444redirects ports 2000 to 2999 (inclusive) to port 4000. 1445rdr ... port 2000:2999 -> ... port 4000:* 1446redirects port 2000 to 4000, 2001 to 4001, ..., 2999 to 4999. 1447.El 1448.Pp 1449In addition to modifying the address, some translation rules may modify 1450source or destination ports for 1451.Xr tcp 4 1452or 1453.Xr udp 4 1454connections; implicitly in the case of 1455.Ar nat 1456rules and both implicitly and explicitly in the case of 1457.Ar rdr 1458rules. 1459A 1460.Ar rdr 1461rule may cause the source port to be modified if doing so avoids a conflict 1462with an existing connection. 1463A random source port in the range 50001-65535 is chosen in this case; to 1464avoid excessive CPU consumption, the number of searches for a free port is 1465limited by the 1466.Va net.pf.rdr_srcport_rewrite_tries 1467sysctl. 1468Port numbers are never translated with a 1469.Ar binat 1470rule. 1471.Pp 1472Evaluation order of the translation rules is dependent on the type 1473of the translation rules and of the direction of a packet. 1474.Ar binat 1475rules are always evaluated first. 1476Then either the 1477.Ar rdr 1478rules are evaluated on an inbound packet or the 1479.Ar nat 1480rules on an outbound packet. 1481Rules of the same type are evaluated in the same order in which they 1482appear in the ruleset. 1483The first matching rule decides what action is taken. 1484.Pp 1485The 1486.Ar no 1487option prefixed to a translation rule causes packets to remain untranslated, 1488much in the same way as 1489.Ar drop quick 1490works in the packet filter (see below). 1491If no rule matches the packet it is passed to the filter engine unmodified. 1492.Pp 1493Translation rules apply only to packets that pass through 1494the specified interface, and if no interface is specified, 1495translation is applied to packets on all interfaces. 1496For instance, redirecting port 80 on an external interface to an internal 1497web server will only work for connections originating from the outside. 1498Connections to the address of the external interface from local hosts will 1499not be redirected, since such packets do not actually pass through the 1500external interface. 1501Redirections cannot reflect packets back through the interface they arrive 1502on, they can only be redirected to hosts connected to different interfaces 1503or to the firewall itself. 1504.Pp 1505Note that redirecting external incoming connections to the loopback 1506address, as in 1507.Bd -literal -offset indent 1508rdr on ne3 inet proto tcp to port smtp -> 127.0.0.1 port spamd 1509.Ed 1510.Pp 1511will effectively allow an external host to connect to daemons 1512bound solely to the loopback address, circumventing the traditional 1513blocking of such connections on a real interface. 1514Unless this effect is desired, any of the local non-loopback addresses 1515should be used as redirection target instead, which allows external 1516connections only to daemons bound to this address or not bound to 1517any address. 1518.Pp 1519See 1520.Sx TRANSLATION EXAMPLES 1521below. 1522.Sh PACKET FILTERING 1523.Xr pf 4 1524has the ability to 1525.Ar block 1526, 1527.Ar pass 1528and 1529.Ar match 1530packets based on attributes of their layer 3 (see 1531.Xr ip 4 1532and 1533.Xr ip6 4 ) 1534and layer 4 (see 1535.Xr icmp 4 , 1536.Xr icmp6 4 , 1537.Xr tcp 4 , 1538.Xr sctp 4 , 1539.Xr udp 4 ) 1540headers. 1541In addition, packets may also be 1542assigned to queues for the purpose of bandwidth control. 1543.Pp 1544For each packet processed by the packet filter, the filter rules are 1545evaluated in sequential order, from first to last. 1546For 1547.Ar block 1548and 1549.Ar pass 1550, the last matching rule decides what action is taken. 1551For 1552.Ar match 1553, rules are evaluated every time they match; the pass/block state of a packet 1554remains unchanged. 1555If no rule matches the packet, the default action is to pass 1556the packet. 1557.Pp 1558The following actions can be used in the filter: 1559.Bl -tag -width xxxx 1560.It Ar block 1561The packet is blocked. 1562There are a number of ways in which a 1563.Ar block 1564rule can behave when blocking a packet. 1565The default behaviour is to 1566.Ar drop 1567packets silently, however this can be overridden or made 1568explicit either globally, by setting the 1569.Ar block-policy 1570option, or on a per-rule basis with one of the following options: 1571.Pp 1572.Bl -tag -width xxxx -compact 1573.It Ar drop 1574The packet is silently dropped. 1575.It Ar return-rst 1576This applies only to 1577.Xr tcp 4 1578packets, and issues a TCP RST which closes the 1579connection. 1580.It Ar return-icmp 1581.It Ar return-icmp6 1582This causes ICMP messages to be returned for packets which match the rule. 1583By default this is an ICMP UNREACHABLE message, however this 1584can be overridden by specifying a message as a code or number. 1585.It Ar return 1586This causes a TCP RST to be returned for 1587.Xr tcp 4 1588packets, an SCTP ABORT for SCTP 1589and an ICMP UNREACHABLE for UDP and other packets. 1590.El 1591.Pp 1592Options returning ICMP packets currently have no effect if 1593.Xr pf 4 1594operates on a 1595.Xr if_bridge 4 , 1596as the code to support this feature has not yet been implemented. 1597.Pp 1598The simplest mechanism to block everything by default and only pass 1599packets that match explicit rules is specify a first filter rule of: 1600.Bd -literal -offset indent 1601block all 1602.Ed 1603.It Ar match 1604The packet is matched. 1605This mechanism is used to provide fine grained filtering without altering the 1606block/pass state of a packet. 1607.Ar match 1608rules differ from 1609.Ar block 1610and 1611.Ar pass 1612rules in that parameters are set for every rule a packet matches, not only 1613on the last matching rule. 1614For the following parameters, this means that the parameter effectively becomes 1615"sticky" until explicitly overridden: 1616.Ar queue , 1617.Ar dnpipe , 1618.Ar dnqueue , 1619.Ar rtable , 1620.Ar scrub 1621. 1622.It Ar pass 1623The packet is passed; 1624state is created unless the 1625.Ar no state 1626option is specified. 1627.El 1628.Pp 1629By default 1630.Xr pf 4 1631filters packets statefully; the first time a packet matches a 1632.Ar pass 1633rule, a state entry is created; for subsequent packets the filter checks 1634whether the packet matches any state. 1635If it does, the packet is passed without evaluation of any rules. 1636After the connection is closed or times out, the state entry is automatically 1637removed. 1638.Pp 1639This has several advantages. 1640For TCP connections, comparing a packet to a state involves checking 1641its sequence numbers, as well as TCP timestamps if a 1642.Ar scrub reassemble tcp 1643rule applies to the connection. 1644If these values are outside the narrow windows of expected 1645values, the packet is dropped. 1646This prevents spoofing attacks, such as when an attacker sends packets with 1647a fake source address/port but does not know the connection's sequence 1648numbers. 1649Similarly, 1650.Xr pf 4 1651knows how to match ICMP replies to states. 1652For example, 1653.Bd -literal -offset indent 1654pass out inet proto icmp all icmp-type echoreq 1655.Ed 1656.Pp 1657allows echo requests (such as those created by 1658.Xr ping 8 ) 1659out statefully, and matches incoming echo replies correctly to states. 1660.Pp 1661Also, looking up states is usually faster than evaluating rules. 1662.Pp 1663Furthermore, correct handling of ICMP error messages is critical to 1664many protocols, particularly TCP. 1665.Xr pf 4 1666matches ICMP error messages to the correct connection, checks them against 1667connection parameters, and passes them if appropriate. 1668For example if an ICMP source quench message referring to a stateful TCP 1669connection arrives, it will be matched to the state and get passed. 1670.Pp 1671Finally, state tracking is required for 1672.Ar nat , binat No and Ar rdr 1673rules, in order to track address and port translations and reverse the 1674translation on returning packets. 1675.Pp 1676.Xr pf 4 1677will also create state for other protocols which are effectively stateless by 1678nature. 1679UDP packets are matched to states using only host addresses and ports, 1680and other protocols are matched to states using only the host addresses. 1681.Pp 1682If stateless filtering of individual packets is desired, 1683the 1684.Ar no state 1685keyword can be used to specify that state will not be created 1686if this is the last matching rule. 1687A number of parameters can also be set to affect how 1688.Xr pf 4 1689handles state tracking. 1690See 1691.Sx STATEFUL TRACKING OPTIONS 1692below for further details. 1693.Sh PARAMETERS 1694The rule parameters specify the packets to which a rule applies. 1695A packet always comes in on, or goes out through, one interface. 1696Most parameters are optional. 1697If a parameter is specified, the rule only applies to packets with 1698matching attributes. 1699Certain parameters can be expressed as lists, in which case 1700.Xr pfctl 8 1701generates all needed rule combinations. 1702.Bl -tag -width xxxx 1703.It Ar in No or Ar out 1704This rule applies to incoming or outgoing packets. 1705If neither 1706.Ar in 1707nor 1708.Ar out 1709are specified, the rule will match packets in both directions. 1710.It Ar log Pq Cm all | matches | to Ao Ar interface Ac | Cm user 1711In addition to any action specified, 1712log the packet. 1713Only the packet that establishes the state is logged, 1714unless the 1715.Ar no state 1716option is specified. 1717The logged packets are sent to a 1718.Xr pflog 4 1719interface, by default pflog0; 1720pflog0 is monitored by the 1721.Xr pflogd 8 1722logging daemon which logs to the file 1723.Pa /var/log/pflog 1724in 1725.Xr pcap 3 1726binary format. 1727.Pp 1728The keywords 1729.Cm all , matches , to , 1730and 1731.Cm user 1732are all optional, 1733and can be combined using commas. 1734.Pp 1735Use 1736.Cm all 1737to force logging of all packets for a connection. 1738This is not necessary when 1739.Ar no state 1740is explicitly specified. 1741.Pp 1742If 1743.Cm matches 1744is specified, 1745it logs the packet on all subsequent matching rules. 1746It is often combined with 1747.Cm to Aq Ar interface 1748to avoid adding noise to the default log file. 1749.Pp 1750The keyword 1751.Cm user 1752logs the 1753.Ux 1754user ID of the user that owns the socket and the PID of the process that 1755has the socket open where the packet is sourced from or destined to 1756(depending on which socket is local). 1757This is in addition to the normal information logged. 1758.Pp 1759Only the first packet 1760logged via 1761.Ar log (all, user) 1762will have the user credentials logged when using stateful matching. 1763.Pp 1764To specify a logging interface other than pflog0, 1765use the syntax 1766.Cm to Aq Ar interface . 1767.It Ar quick 1768If a packet matches a rule which has the 1769.Ar quick 1770option set, this rule 1771is considered the last matching rule, and evaluation of subsequent rules 1772is skipped. 1773.It Ar on Aq Ar interface 1774This rule applies only to packets coming in on, or going out through, this 1775particular interface or interface group. 1776For more information on interface groups, 1777see the 1778.Ic group 1779keyword in 1780.Xr ifconfig 8 . 1781.Ar any 1782will match any existing interface except loopback ones. 1783.It Aq Ar af 1784This rule applies only to packets of this address family. 1785Supported values are 1786.Ar inet 1787and 1788.Ar inet6 . 1789.It Ar proto Aq Ar protocol 1790This rule applies only to packets of this protocol. 1791Common protocols are 1792.Xr icmp 4 , 1793.Xr icmp6 4 , 1794.Xr tcp 4 , 1795.Xr sctp 4 , 1796and 1797.Xr udp 4 . 1798For a list of all the protocol name to number mappings used by 1799.Xr pfctl 8 , 1800see the file 1801.Pa /etc/protocols . 1802.It Xo 1803.Ar from Aq Ar source 1804.Ar port Aq Ar source 1805.Ar os Aq Ar source 1806.Ar to Aq Ar dest 1807.Ar port Aq Ar dest 1808.Xc 1809This rule applies only to packets with the specified source and destination 1810addresses and ports. 1811.Pp 1812Addresses can be specified in CIDR notation (matching netblocks), as 1813symbolic host names, interface names or interface group names, or as any 1814of the following keywords: 1815.Pp 1816.Bl -tag -width xxxxxxxxxxxxxx -compact 1817.It Ar any 1818Any address. 1819.It Ar no-route 1820Any address which is not currently routable. 1821.It Ar urpf-failed 1822Any source address that fails a unicast reverse path forwarding (URPF) 1823check, i.e. packets coming in on an interface other than that which holds 1824the route back to the packet's source address. 1825.It Ar self 1826Expands to all addresses assigned to all interfaces. 1827.It Aq Ar table 1828Any address that matches the given table. 1829.El 1830.Pp 1831Ranges of addresses are specified by using the 1832.Sq - 1833operator. 1834For instance: 1835.Dq 10.1.1.10 - 10.1.1.12 1836means all addresses from 10.1.1.10 to 10.1.1.12, 1837hence addresses 10.1.1.10, 10.1.1.11, and 10.1.1.12. 1838.Pp 1839Interface names and interface group names, and 1840.Ar self 1841can have modifiers appended: 1842.Pp 1843.Bl -tag -width xxxxxxxxxxxx -compact 1844.It Ar :network 1845Translates to the network(s) attached to the interface. 1846.It Ar :broadcast 1847Translates to the interface's broadcast address(es). 1848.It Ar :peer 1849Translates to the point-to-point interface's peer address(es). 1850.It Ar :0 1851Do not include interface aliases. 1852.El 1853.Pp 1854Host names may also have the 1855.Ar :0 1856option appended to restrict the name resolution to the first of each 1857v4 and non-link-local v6 address found. 1858.Pp 1859Host name resolution and interface to address translation are done at 1860ruleset load-time. 1861When the address of an interface (or host name) changes (under DHCP or PPP, 1862for instance), the ruleset must be reloaded for the change to be reflected 1863in the kernel. 1864Surrounding the interface name (and optional modifiers) in parentheses 1865changes this behaviour. 1866When the interface name is surrounded by parentheses, the rule is 1867automatically updated whenever the interface changes its address. 1868The ruleset does not need to be reloaded. 1869This is especially useful with 1870.Ar nat . 1871.Pp 1872Ports can be specified either by number or by name. 1873For example, port 80 can be specified as 1874.Em www . 1875For a list of all port name to number mappings used by 1876.Xr pfctl 8 , 1877see the file 1878.Pa /etc/services . 1879.Pp 1880Ports and ranges of ports are specified by using these operators: 1881.Bd -literal -offset indent 1882= (equal) 1883!= (unequal) 1884< (less than) 1885<= (less than or equal) 1886> (greater than) 1887>= (greater than or equal) 1888: (range including boundaries) 1889>< (range excluding boundaries) 1890<> (except range) 1891.Ed 1892.Pp 1893.Sq >< , 1894.Sq <> 1895and 1896.Sq \&: 1897are binary operators (they take two arguments). 1898For instance: 1899.Bl -tag -width Fl 1900.It Ar port 2000:2004 1901means 1902.Sq all ports >= 2000 and <= 2004 , 1903hence ports 2000, 2001, 2002, 2003 and 2004. 1904.It Ar port 2000 >< 2004 1905means 1906.Sq all ports > 2000 and < 2004 , 1907hence ports 2001, 2002 and 2003. 1908.It Ar port 2000 <> 2004 1909means 1910.Sq all ports < 2000 or > 2004 , 1911hence ports 1-1999 and 2005-65535. 1912.El 1913.Pp 1914The operating system of the source host can be specified in the case of TCP 1915rules with the 1916.Ar OS 1917modifier. 1918See the 1919.Sx OPERATING SYSTEM FINGERPRINTING 1920section for more information. 1921.Pp 1922The host, port and OS specifications are optional, as in the following examples: 1923.Bd -literal -offset indent 1924pass in all 1925pass in from any to any 1926pass in proto tcp from any port < 1024 to any 1927pass in proto tcp from any to any port 25 1928pass in proto tcp from 10.0.0.0/8 port >= 1024 \e 1929 to ! 10.1.2.3 port != ssh 1930pass in proto tcp from any os "OpenBSD" 1931.Ed 1932.It Ar all 1933This is equivalent to "from any to any". 1934.It Ar group Aq Ar group 1935Similar to 1936.Ar user , 1937this rule only applies to packets of sockets owned by the specified group. 1938.It Ar user Aq Ar user 1939This rule only applies to packets of sockets owned by the specified user. 1940For outgoing connections initiated from the firewall, this is the user 1941that opened the connection. 1942For incoming connections to the firewall itself, this is the user that 1943listens on the destination port. 1944For forwarded connections, where the firewall is not a connection endpoint, 1945the user and group are 1946.Em unknown . 1947.Pp 1948All packets, both outgoing and incoming, of one connection are associated 1949with the same user and group. 1950Only TCP and UDP packets can be associated with users; for other protocols 1951these parameters are ignored. 1952.Pp 1953User and group refer to the effective (as opposed to the real) IDs, in 1954case the socket is created by a setuid/setgid process. 1955User and group IDs are stored when a socket is created; 1956when a process creates a listening socket as root (for instance, by 1957binding to a privileged port) and subsequently changes to another 1958user ID (to drop privileges), the credentials will remain root. 1959.Pp 1960User and group IDs can be specified as either numbers or names. 1961The syntax is similar to the one for ports. 1962The value 1963.Em unknown 1964matches packets of forwarded connections. 1965.Em unknown 1966can only be used with the operators 1967.Cm = 1968and 1969.Cm != . 1970Other constructs like 1971.Cm user \*(Ge unknown 1972are invalid. 1973Forwarded packets with unknown user and group ID match only rules 1974that explicitly compare against 1975.Em unknown 1976with the operators 1977.Cm = 1978or 1979.Cm != . 1980For instance 1981.Cm user \*(Ge 0 1982does not match forwarded packets. 1983The following example allows only selected users to open outgoing 1984connections: 1985.Bd -literal -offset indent 1986block out proto { tcp, udp } all 1987pass out proto { tcp, udp } all user { < 1000, dhartmei } 1988.Ed 1989.It Xo Ar flags Aq Ar a 1990.Pf / Ns Aq Ar b 1991.No \*(Ba / Ns Aq Ar b 1992.No \*(Ba any 1993.Xc 1994This rule only applies to TCP packets that have the flags 1995.Aq Ar a 1996set out of set 1997.Aq Ar b . 1998Flags not specified in 1999.Aq Ar b 2000are ignored. 2001For stateful connections, the default is 2002.Ar flags S/SA . 2003To indicate that flags should not be checked at all, specify 2004.Ar flags any . 2005The flags are: (F)IN, (S)YN, (R)ST, (P)USH, (A)CK, (U)RG, (E)CE, and C(W)R. 2006.Bl -tag -width Fl 2007.It Ar flags S/S 2008Flag SYN is set. 2009The other flags are ignored. 2010.It Ar flags S/SA 2011This is the default setting for stateful connections. 2012Out of SYN and ACK, exactly SYN may be set. 2013SYN, SYN+PSH and SYN+RST match, but SYN+ACK, ACK and ACK+RST do not. 2014This is more restrictive than the previous example. 2015.It Ar flags /SFRA 2016If the first set is not specified, it defaults to none. 2017All of SYN, FIN, RST and ACK must be unset. 2018.El 2019.Pp 2020Because 2021.Ar flags S/SA 2022is applied by default (unless 2023.Ar no state 2024is specified), only the initial SYN packet of a TCP handshake will create 2025a state for a TCP connection. 2026It is possible to be less restrictive, and allow state creation from 2027intermediate 2028.Pq non-SYN 2029packets, by specifying 2030.Ar flags any . 2031This will cause 2032.Xr pf 4 2033to synchronize to existing connections, for instance 2034if one flushes the state table. 2035However, states created from such intermediate packets may be missing 2036connection details such as the TCP window scaling factor. 2037States which modify the packet flow, such as those affected by 2038.Ar af-to, 2039.Ar nat, 2040.Ar binat or 2041.Ar rdr 2042rules, 2043.Ar modulate No or Ar synproxy state 2044options, or scrubbed with 2045.Ar reassemble tcp 2046will also not be recoverable from intermediate packets. 2047Such connections will stall and time out. 2048.It Xo Ar icmp-type Aq Ar type 2049.Ar code Aq Ar code 2050.Xc 2051.It Xo Ar icmp6-type Aq Ar type 2052.Ar code Aq Ar code 2053.Xc 2054This rule only applies to ICMP or ICMPv6 packets with the specified type 2055and code. 2056Text names for ICMP types and codes are listed in 2057.Xr icmp 4 2058and 2059.Xr icmp6 4 . 2060This parameter is only valid for rules that cover protocols ICMP or 2061ICMP6. 2062The protocol and the ICMP type indicator 2063.Po 2064.Ar icmp-type 2065or 2066.Ar icmp6-type 2067.Pc 2068must match. 2069.It Xo Ar tos Aq Ar string 2070.No \*(Ba Aq Ar number 2071.Xc 2072This rule applies to packets with the specified 2073.Em TOS 2074bits set. 2075.Em TOS 2076may be 2077given as one of 2078.Ar critical , 2079.Ar inetcontrol , 2080.Ar lowdelay , 2081.Ar netcontrol , 2082.Ar throughput , 2083.Ar reliability , 2084or one of the DiffServ Code Points: 2085.Ar ef , 2086.Ar va , 2087.Ar af11 No ... Ar af43 , 2088.Ar cs0 No ... Ar cs7 ; 2089or as either hex or decimal. 2090.Pp 2091For example, the following rules are identical: 2092.Bd -literal -offset indent 2093pass all tos lowdelay 2094pass all tos 0x10 2095pass all tos 16 2096.Ed 2097.It Ar allow-opts 2098By default, IPv4 packets with IP options or IPv6 packets with routing 2099extension headers are blocked. 2100When 2101.Ar allow-opts 2102is specified for a 2103.Ar pass 2104rule, packets that pass the filter based on that rule (last matching) 2105do so even if they contain IP options or routing extension headers. 2106For packets that match state, the rule that initially created the 2107state is used. 2108The implicit 2109.Ar pass 2110rule that is used when a packet does not match any rules does not 2111allow IP options. 2112.It Ar label Aq Ar string 2113Adds a label (name) to the rule, which can be used to identify the rule. 2114For instance, 2115pfctl -s labels 2116shows per-rule statistics for rules that have labels. 2117.Pp 2118The following macros can be used in labels: 2119.Pp 2120.Bl -tag -width $srcaddr -compact -offset indent 2121.It Ar $if 2122The interface. 2123.It Ar $srcaddr 2124The source IP address. 2125.It Ar $dstaddr 2126The destination IP address. 2127.It Ar $srcport 2128The source port specification. 2129.It Ar $dstport 2130The destination port specification. 2131.It Ar $proto 2132The protocol name. 2133.It Ar $nr 2134The rule number. 2135.El 2136.Pp 2137For example: 2138.Bd -literal -offset indent 2139ips = \&"{ 1.2.3.4, 1.2.3.5 }\&" 2140pass in proto tcp from any to $ips \e 2141 port > 1023 label \&"$dstaddr:$dstport\&" 2142.Ed 2143.Pp 2144expands to 2145.Bd -literal -offset indent 2146pass in inet proto tcp from any to 1.2.3.4 \e 2147 port > 1023 label \&"1.2.3.4:>1023\&" 2148pass in inet proto tcp from any to 1.2.3.5 \e 2149 port > 1023 label \&"1.2.3.5:>1023\&" 2150.Ed 2151.Pp 2152The macro expansion for the 2153.Ar label 2154directive occurs only at configuration file parse time, not during runtime. 2155.It Ar ridentifier Aq Ar number 2156Add an identifier (number) to the rule, which can be used to correlate the rule 2157to pflog entries, even after ruleset updates. 2158.It Xo Ar queue Aq Ar queue 2159.No \*(Ba ( Aq Ar queue , 2160.Aq Ar queue ) 2161.Xc 2162Packets matching this rule will be assigned to the specified queue. 2163If two queues are given, packets which have a 2164.Em TOS 2165of 2166.Em lowdelay 2167and TCP ACKs with no data payload will be assigned to the second one. 2168See 2169.Sx QUEUEING 2170for setup details. 2171.Pp 2172For example: 2173.Bd -literal -offset indent 2174pass in proto tcp to port 25 queue mail 2175pass in proto tcp to port 22 queue(ssh_bulk, ssh_prio) 2176.Ed 2177.It Cm set prio Ar priority | Pq Ar priority , priority 2178Packets matching this rule will be assigned a specific queueing priority. 2179Priorities are assigned as integers 0 through 7. 2180If the packet is transmitted on a 2181.Xr vlan 4 2182interface, the queueing priority will be written as the priority 2183code point in the 802.1Q VLAN header. 2184If two priorities are given, packets which have a TOS of 2185.Cm lowdelay 2186and TCP ACKs with no data payload will be assigned to the second one. 2187.Pp 2188For example: 2189.Bd -literal -offset indent 2190pass in proto tcp to port 25 set prio 2 2191pass in proto tcp to port 22 set prio (2, 5) 2192.Ed 2193.It Ar received-on Aq Ar interface 2194Only match packets which were received on the specified 2195.Ar interface 2196(or interface group). 2197.Ar any 2198will match any existing interface except loopback ones. 2199.It Ar tag Aq Ar string 2200Packets matching this rule will be tagged with the 2201specified string. 2202The tag acts as an internal marker that can be used to 2203identify these packets later on. 2204This can be used, for example, to provide trust between 2205interfaces and to determine if packets have been 2206processed by translation rules. 2207Tags are 2208.Qq sticky , 2209meaning that the packet will be tagged even if the rule 2210is not the last matching rule. 2211Further matching rules can replace the tag with a 2212new one but will not remove a previously applied tag. 2213A packet is only ever assigned one tag at a time. 2214Packet tagging can be done during 2215.Ar nat , 2216.Ar rdr , 2217.Ar binat 2218or 2219.Ar ether 2220rules in addition to filter rules. 2221Tags take the same macros as labels (see above). 2222.It Ar tagged Aq Ar string 2223Used with filter, translation or scrub rules 2224to specify that packets must already 2225be tagged with the given tag in order to match the rule. 2226Inverse tag matching can also be done 2227by specifying the 2228.Cm !\& 2229operator before the 2230.Ar tagged 2231keyword. 2232.It Ar rtable Aq Ar number 2233Used to select an alternate routing table for the routing lookup. 2234Only effective before the route lookup happened, i.e. when filtering inbound. 2235.It Xo Ar divert-to Aq Ar host 2236.Ar port Aq Ar port 2237.Xc 2238Used to 2239.Xr divert 4 2240packets to the given divert 2241.Ar port . 2242Historically 2243.Ox pf has another meaning for this, and 2244.Fx pf uses 2245this syntax to support 2246.Xr divert 4 instead. Hence, 2247.Ar host 2248has no meaning and can be set to anything like 127.0.0.1. 2249If a packet is re-injected and does not change direction then it will not be 2250re-diverted. 2251.It Ar divert-reply 2252It has no meaning in 2253.Fx pf . 2254.It Ar probability Aq Ar number 2255A probability attribute can be attached to a rule, with a value set between 22560 and 1, bounds not included. 2257In that case, the rule will be honoured using the given probability value 2258only. 2259For example, the following rule will drop 20% of incoming ICMP packets: 2260.Bd -literal -offset indent 2261block in proto icmp probability 20% 2262.Ed 2263.It Ar prio Aq Ar number 2264Only match packets which have the given queueing priority assigned. 2265.El 2266.Sh ROUTING 2267If a packet matches a rule with a route option set, the packet filter will 2268route the packet according to the type of route option. 2269When such a rule creates state, the route option is also applied to all 2270packets matching the same connection. 2271.Bl -tag -width xxxx 2272.It Ar route-to 2273The 2274.Ar route-to 2275option routes the packet to the specified interface with an optional address 2276for the next hop. 2277When a 2278.Ar route-to 2279rule creates state, only packets that pass in the same direction as the 2280filter rule specifies will be routed in this way. 2281Packets passing in the opposite direction (replies) are not affected 2282and are routed normally. 2283.It Ar reply-to 2284The 2285.Ar reply-to 2286option is similar to 2287.Ar route-to , 2288but routes packets that pass in the opposite direction (replies) to the 2289specified interface. 2290Opposite direction is only defined in the context of a state entry, and 2291.Ar reply-to 2292is useful only in rules that create state. 2293It can be used on systems with multiple external connections to 2294route all outgoing packets of a connection through the interface 2295the incoming connection arrived through (symmetric routing enforcement). 2296.It Ar dup-to 2297The 2298.Ar dup-to 2299option creates a duplicate of the packet and routes it like 2300.Ar route-to . 2301The original packet gets routed as it normally would. 2302.El 2303.Sh POOL OPTIONS 2304For 2305.Ar nat 2306and 2307.Ar rdr 2308rules, (as well as for the 2309.Ar route-to , 2310.Ar reply-to 2311and 2312.Ar dup-to 2313rule options) for which there is a single redirection address which has a 2314subnet mask smaller than 32 for IPv4 or 128 for IPv6 (more than one IP 2315address), a variety of different methods for assigning this address can be 2316used: 2317.Bl -tag -width xxxx 2318.It Ar bitmask 2319The 2320.Ar bitmask 2321option applies the network portion of the redirection address to the address 2322to be modified (source with 2323.Ar nat , 2324destination with 2325.Ar rdr ) . 2326.It Ar random 2327The 2328.Ar random 2329option selects an address at random within the defined block of addresses. 2330.It Ar source-hash 2331The 2332.Ar source-hash 2333option uses a hash of the source address to determine the redirection address, 2334ensuring that the redirection address is always the same for a given source. 2335An optional key can be specified after this keyword either in hex or as a 2336string; by default 2337.Xr pfctl 8 2338randomly generates a key for source-hash every time the 2339ruleset is reloaded. 2340.It Ar round-robin 2341The 2342.Ar round-robin 2343option loops through the redirection address(es). 2344.Pp 2345When more than one redirection address is specified, 2346.Ar bitmask 2347is not permitted as a pool type. 2348.It Ar static-port 2349With 2350.Ar nat 2351rules, the 2352.Ar static-port 2353option prevents 2354.Xr pf 4 2355from modifying the source port on TCP and UDP packets. 2356.It Xo Ar map-e-portset Aq Ar psid-offset 2357.No / Aq Ar psid-len 2358.No / Aq Ar psid 2359.Xc 2360With 2361.Ar nat 2362rules, the 2363.Ar map-e-portset 2364option enables the source port translation of MAP-E (RFC 7597) Customer Edge. 2365In order to make the host act as a MAP-E Customer Edge, setting up a tunneling 2366interface and pass rules for encapsulated packets are required in addition 2367to the map-e-portset nat rule. 2368.Pp 2369For example: 2370.Bd -literal -offset indent 2371nat on $gif_mape_if from $int_if:network to any \e 2372 -> $ipv4_mape_src map-e-portset 6/8/0x34 2373.Ed 2374.Pp 2375sets PSID offset 6, PSID length 8, PSID 0x34. 2376.It Ar endpoint-independent 2377With 2378.Ar nat 2379rules, the 2380.Ar endpoint-independent 2381option caues 2382.Xr pf 4 2383to always map connections from a UDP source address and port to the same 2384NAT address and port. 2385This feature implements "full-cone" NAT behavior. 2386.El 2387.Pp 2388Additionally, the 2389.Ar sticky-address 2390option can be specified to help ensure that multiple connections from the 2391same source are mapped to the same redirection address. 2392This option can be used with the 2393.Ar random 2394and 2395.Ar round-robin 2396pool options. 2397Note that by default these associations are destroyed as soon as there are 2398no longer states which refer to them; in order to make the mappings last 2399beyond the lifetime of the states, increase the global options with 2400.Ar set timeout src.track . 2401See 2402.Sx STATEFUL TRACKING OPTIONS 2403for more ways to control the source tracking. 2404.Sh STATE MODULATION 2405Much of the security derived from TCP is attributable to how well the 2406initial sequence numbers (ISNs) are chosen. 2407Some popular stack implementations choose 2408.Em very 2409poor ISNs and thus are normally susceptible to ISN prediction exploits. 2410By applying a 2411.Ar modulate state 2412rule to a TCP connection, 2413.Xr pf 4 2414will create a high quality random sequence number for each connection 2415endpoint. 2416.Pp 2417The 2418.Ar modulate state 2419directive implicitly keeps state on the rule and is 2420only applicable to TCP connections. 2421.Pp 2422For instance: 2423.Bd -literal -offset indent 2424block all 2425pass out proto tcp from any to any modulate state 2426pass in proto tcp from any to any port 25 flags S/SFRA modulate state 2427.Ed 2428.Pp 2429Note that modulated connections will not recover when the state table 2430is lost (firewall reboot, flushing the state table, etc...). 2431.Xr pf 4 2432will not be able to infer a connection again after the state table flushes 2433the connection's modulator. 2434When the state is lost, the connection may be left dangling until the 2435respective endpoints time out the connection. 2436It is possible on a fast local network for the endpoints to start an ACK 2437storm while trying to resynchronize after the loss of the modulator. 2438The default 2439.Ar flags 2440settings (or a more strict equivalent) should be used on 2441.Ar modulate state 2442rules to prevent ACK storms. 2443.Pp 2444Note that alternative methods are available 2445to prevent loss of the state table 2446and allow for firewall failover. 2447See 2448.Xr carp 4 2449and 2450.Xr pfsync 4 2451for further information. 2452.Sh SYN PROXY 2453By default, 2454.Xr pf 4 2455passes packets that are part of a 2456.Xr tcp 4 2457handshake between the endpoints. 2458The 2459.Ar synproxy state 2460option can be used to cause 2461.Xr pf 4 2462itself to complete the handshake with the active endpoint, perform a handshake 2463with the passive endpoint, and then forward packets between the endpoints. 2464.Pp 2465No packets are sent to the passive endpoint before the active endpoint has 2466completed the handshake, hence so-called SYN floods with spoofed source 2467addresses will not reach the passive endpoint, as the sender can't complete the 2468handshake. 2469.Pp 2470The proxy is transparent to both endpoints, they each see a single 2471connection from/to the other endpoint. 2472.Xr pf 4 2473chooses random initial sequence numbers for both handshakes. 2474Once the handshakes are completed, the sequence number modulators 2475(see previous section) are used to translate further packets of the 2476connection. 2477.Ar synproxy state 2478includes 2479.Ar modulate state . 2480.Pp 2481Rules with 2482.Ar synproxy 2483will not work if 2484.Xr pf 4 2485operates on a 2486.Xr bridge 4 . 2487.Pp 2488Example: 2489.Bd -literal -offset indent 2490pass in proto tcp from any to any port www synproxy state 2491.Ed 2492.Sh STATEFUL TRACKING OPTIONS 2493A number of options related to stateful tracking can be applied on a 2494per-rule basis. 2495.Ar keep state , 2496.Ar modulate state 2497and 2498.Ar synproxy state 2499support these options, and 2500.Ar keep state 2501must be specified explicitly to apply options to a rule. 2502.Pp 2503.Bl -tag -width xxxx -compact 2504.It Ar max Aq Ar number 2505Limits the number of concurrent states the rule may create. 2506When this limit is reached, further packets that would create 2507state are dropped until existing states time out. 2508.It Ar no-sync 2509Prevent state changes for states created by this rule from appearing on the 2510.Xr pfsync 4 2511interface. 2512.It Xo Aq Ar timeout 2513.Aq Ar seconds 2514.Xc 2515Changes the timeout values used for states created by this rule. 2516For a list of all valid timeout names, see 2517.Sx OPTIONS 2518above. 2519.It Ar sloppy 2520Uses a sloppy TCP connection tracker that does not check sequence 2521numbers at all, which makes insertion and ICMP teardown attacks way 2522easier. 2523This is intended to be used in situations where one does not see all 2524packets of a connection, e.g. in asymmetric routing situations. 2525Cannot be used with modulate or synproxy state. 2526.It Ar pflow 2527States created by this rule are exported on the 2528.Xr pflow 4 2529interface. 2530.It Ar allow-related 2531Automatically allow connections related to this one, regardless of rules that 2532might otherwise affect them. 2533This currently only applies to SCTP multihomed connection. 2534.El 2535.Pp 2536Multiple options can be specified, separated by commas: 2537.Bd -literal -offset indent 2538pass in proto tcp from any to any \e 2539 port www keep state \e 2540 (max 100, source-track rule, max-src-nodes 75, \e 2541 max-src-states 3, tcp.established 60, tcp.closing 5) 2542.Ed 2543.Pp 2544When the 2545.Ar source-track 2546keyword is specified, the number of states per source IP is tracked. 2547.Pp 2548.Bl -tag -width xxxx -compact 2549.It Ar source-track rule 2550The maximum number of states created by this rule is limited by the rule's 2551.Ar max-src-nodes 2552and 2553.Ar max-src-states 2554options. 2555Only state entries created by this particular rule count toward the rule's 2556limits. 2557.It Ar source-track global 2558The number of states created by all rules that use this option is limited. 2559Each rule can specify different 2560.Ar max-src-nodes 2561and 2562.Ar max-src-states 2563options, however state entries created by any participating rule count towards 2564each individual rule's limits. 2565.El 2566.Pp 2567The following limits can be set: 2568.Pp 2569.Bl -tag -width xxxx -compact 2570.It Ar max-src-nodes Aq Ar number 2571Limits the maximum number of source addresses which can simultaneously 2572have state table entries. 2573.It Ar max-src-states Aq Ar number 2574Limits the maximum number of simultaneous state entries that a single 2575source address can create with this rule. 2576.El 2577.Pp 2578For stateful TCP connections, limits on established connections (connections 2579which have completed the TCP 3-way handshake) can also be enforced 2580per source IP. 2581.Pp 2582.Bl -tag -width xxxx -compact 2583.It Ar max-src-conn Aq Ar number 2584Limits the maximum number of simultaneous TCP connections which have 2585completed the 3-way handshake that a single host can make. 2586.It Xo Ar max-src-conn-rate Aq Ar number 2587.No / Aq Ar seconds 2588.Xc 2589Limit the rate of new connections over a time interval. 2590The connection rate is an approximation calculated as a moving average. 2591.El 2592.Pp 2593When one of these limits is reached, further packets that would create 2594state are dropped until existing states time out. 2595.Pp 2596Because the 3-way handshake ensures that the source address is not being 2597spoofed, more aggressive action can be taken based on these limits. 2598With the 2599.Ar overload Aq Ar table 2600state option, source IP addresses which hit either of the limits on 2601established connections will be added to the named table. 2602This table can be used in the ruleset to block further activity from 2603the offending host, redirect it to a tarpit process, or restrict its 2604bandwidth. 2605.Pp 2606The optional 2607.Ar flush 2608keyword kills all states created by the matching rule which originate 2609from the host which exceeds these limits. 2610The 2611.Ar global 2612modifier to the flush command kills all states originating from the 2613offending host, regardless of which rule created the state. 2614.Pp 2615For example, the following rules will protect the webserver against 2616hosts making more than 100 connections in 10 seconds. 2617Any host which connects faster than this rate will have its address added 2618to the 2619.Aq bad_hosts 2620table and have all states originating from it flushed. 2621Any new packets arriving from this host will be dropped unconditionally 2622by the block rule. 2623.Bd -literal -offset indent 2624block quick from <bad_hosts> 2625pass in on $ext_if proto tcp to $webserver port www keep state \e 2626 (max-src-conn-rate 100/10, overload <bad_hosts> flush global) 2627.Ed 2628.Sh OPERATING SYSTEM FINGERPRINTING 2629Passive OS Fingerprinting is a mechanism to inspect nuances of a TCP 2630connection's initial SYN packet and guess at the host's operating system. 2631Unfortunately these nuances are easily spoofed by an attacker so the 2632fingerprint is not useful in making security decisions. 2633But the fingerprint is typically accurate enough to make policy decisions 2634upon. 2635.Pp 2636The fingerprints may be specified by operating system class, by 2637version, or by subtype/patchlevel. 2638The class of an operating system is typically the vendor or genre 2639and would be 2640.Ox 2641for the 2642.Xr pf 4 2643firewall itself. 2644The version of the oldest available 2645.Ox 2646release on the main FTP site 2647would be 2.6 and the fingerprint would be written 2648.Pp 2649.Dl \&"OpenBSD 2.6\&" 2650.Pp 2651The subtype of an operating system is typically used to describe the 2652patchlevel if that patch led to changes in the TCP stack behavior. 2653In the case of 2654.Ox , 2655the only subtype is for a fingerprint that was 2656normalized by the 2657.Ar no-df 2658scrub option and would be specified as 2659.Pp 2660.Dl \&"OpenBSD 3.3 no-df\&" 2661.Pp 2662Fingerprints for most popular operating systems are provided by 2663.Xr pf.os 5 . 2664Once 2665.Xr pf 4 2666is running, a complete list of known operating system fingerprints may 2667be listed by running: 2668.Pp 2669.Dl # pfctl -so 2670.Pp 2671Filter rules can enforce policy at any level of operating system specification 2672assuming a fingerprint is present. 2673Policy could limit traffic to approved operating systems or even ban traffic 2674from hosts that aren't at the latest service pack. 2675.Pp 2676The 2677.Ar unknown 2678class can also be used as the fingerprint which will match packets for 2679which no operating system fingerprint is known. 2680.Pp 2681Examples: 2682.Bd -literal -offset indent 2683pass out proto tcp from any os OpenBSD 2684block out proto tcp from any os Doors 2685block out proto tcp from any os "Doors PT" 2686block out proto tcp from any os "Doors PT SP3" 2687block out from any os "unknown" 2688pass on lo0 proto tcp from any os "OpenBSD 3.3 lo0" 2689.Ed 2690.Pp 2691Operating system fingerprinting is limited only to the TCP SYN packet. 2692This means that it will not work on other protocols and will not match 2693a currently established connection. 2694.Pp 2695Caveat: operating system fingerprints are occasionally wrong. 2696There are three problems: an attacker can trivially craft his packets to 2697appear as any operating system he chooses; 2698an operating system patch could change the stack behavior and no fingerprints 2699will match it until the database is updated; 2700and multiple operating systems may have the same fingerprint. 2701.Sh BLOCKING SPOOFED TRAFFIC 2702"Spoofing" is the faking of IP addresses, typically for malicious 2703purposes. 2704The 2705.Ar antispoof 2706directive expands to a set of filter rules which will block all 2707traffic with a source IP from the network(s) directly connected 2708to the specified interface(s) from entering the system through 2709any other interface. 2710.Pp 2711For example, the line 2712.Bd -literal -offset indent 2713antispoof for lo0 2714.Ed 2715.Pp 2716expands to 2717.Bd -literal -offset indent 2718block drop in on ! lo0 inet from 127.0.0.1/8 to any 2719block drop in on ! lo0 inet6 from ::1 to any 2720.Ed 2721.Pp 2722For non-loopback interfaces, there are additional rules to block incoming 2723packets with a source IP address identical to the interface's IP(s). 2724For example, assuming the interface wi0 had an IP address of 10.0.0.1 and a 2725netmask of 255.255.255.0, 2726the line 2727.Bd -literal -offset indent 2728antispoof for wi0 inet 2729.Ed 2730.Pp 2731expands to 2732.Bd -literal -offset indent 2733block drop in on ! wi0 inet from 10.0.0.0/24 to any 2734block drop in inet from 10.0.0.1 to any 2735.Ed 2736.Pp 2737Caveat: Rules created by the 2738.Ar antispoof 2739directive interfere with packets sent over loopback interfaces 2740to local addresses. 2741One should pass these explicitly. 2742.Sh FRAGMENT HANDLING 2743The size of IP datagrams (packets) can be significantly larger than the 2744maximum transmission unit (MTU) of the network. 2745In cases when it is necessary or more efficient to send such large packets, 2746the large packet will be fragmented into many smaller packets that will each 2747fit onto the wire. 2748Unfortunately for a firewalling device, only the first logical fragment will 2749contain the necessary header information for the subprotocol that allows 2750.Xr pf 4 2751to filter on things such as TCP ports or to perform NAT. 2752.Pp 2753Besides the use of 2754.Ar set reassemble 2755option or 2756.Ar scrub 2757rules as described in 2758.Sx TRAFFIC NORMALIZATION 2759above, there are three options for handling fragments in the packet filter. 2760.Pp 2761One alternative is to filter individual fragments with filter rules. 2762If no 2763.Ar scrub 2764rule applies to a fragment or 2765.Ar set reassemble 2766is set to 2767.Cm no 2768, it is passed to the filter. 2769Filter rules with matching IP header parameters decide whether the 2770fragment is passed or blocked, in the same way as complete packets 2771are filtered. 2772Without reassembly, fragments can only be filtered based on IP header 2773fields (source/destination address, protocol), since subprotocol header 2774fields are not available (TCP/UDP port numbers, ICMP code/type). 2775The 2776.Ar fragment 2777option can be used to restrict filter rules to apply only to 2778fragments, but not complete packets. 2779Filter rules without the 2780.Ar fragment 2781option still apply to fragments, if they only specify IP header fields. 2782For instance, the rule 2783.Bd -literal -offset indent 2784pass in proto tcp from any to any port 80 2785.Ed 2786.Pp 2787never applies to a fragment, even if the fragment is part of a TCP 2788packet with destination port 80, because without reassembly this information 2789is not available for each fragment. 2790This also means that fragments cannot create new or match existing 2791state table entries, which makes stateful filtering and address 2792translation (NAT, redirection) for fragments impossible. 2793.Pp 2794It's also possible to reassemble only certain fragments by specifying 2795source or destination addresses or protocols as parameters in 2796.Ar scrub 2797rules. 2798.Pp 2799In most cases, the benefits of reassembly outweigh the additional 2800memory cost, and it's recommended to use 2801.Ar set reassemble 2802option or 2803.Ar scrub 2804rules with the 2805.Ar fragment reassemble 2806modifier to reassemble 2807all fragments. 2808.Pp 2809The memory allocated for fragment caching can be limited using 2810.Xr pfctl 8 . 2811Once this limit is reached, fragments that would have to be cached 2812are dropped until other entries time out. 2813The timeout value can also be adjusted. 2814.Pp 2815When forwarding reassembled IPv6 packets, pf refragments them with 2816the original maximum fragment size. 2817This allows the sender to determine the optimal fragment size by 2818path MTU discovery. 2819.Sh ANCHORS 2820Besides the main ruleset, 2821.Xr pfctl 8 2822can load rulesets into 2823.Ar anchor 2824attachment points. 2825An 2826.Ar anchor 2827is a container that can hold rules, address tables, and other anchors. 2828.Pp 2829An 2830.Ar anchor 2831has a name which specifies the path where 2832.Xr pfctl 8 2833can be used to access the anchor to perform operations on it, such as 2834attaching child anchors to it or loading rules into it. 2835Anchors may be nested, with components separated by 2836.Sq / 2837characters, similar to how file system hierarchies are laid out. 2838The main ruleset is actually the default anchor, so filter and 2839translation rules, for example, may also be contained in any anchor. 2840.Pp 2841An anchor can reference another 2842.Ar anchor 2843attachment point 2844using the following kinds 2845of rules: 2846.Bl -tag -width xxxx 2847.It Ar nat-anchor Aq Ar name 2848Evaluates the 2849.Ar nat 2850rules in the specified 2851.Ar anchor . 2852.It Ar rdr-anchor Aq Ar name 2853Evaluates the 2854.Ar rdr 2855rules in the specified 2856.Ar anchor . 2857.It Ar binat-anchor Aq Ar name 2858Evaluates the 2859.Ar binat 2860rules in the specified 2861.Ar anchor . 2862.It Ar anchor Aq Ar name 2863Evaluates the filter rules in the specified 2864.Ar anchor . 2865.It Xo Ar load anchor 2866.Aq Ar name 2867.Ar from Aq Ar file 2868.Xc 2869Loads the rules from the specified file into the 2870anchor 2871.Ar name . 2872.El 2873.Pp 2874When evaluation of the main ruleset reaches an 2875.Ar anchor 2876rule, 2877.Xr pf 4 2878will proceed to evaluate all rules specified in that anchor. 2879.Pp 2880Matching filter and translation rules marked with the 2881.Ar quick 2882option are final and abort the evaluation of the rules in other 2883anchors and the main ruleset. 2884If the 2885.Ar anchor 2886itself is marked with the 2887.Ar quick 2888option, 2889ruleset evaluation will terminate when the anchor is exited if the packet is 2890matched by any rule within the anchor. 2891.Pp 2892.Ar anchor 2893rules are evaluated relative to the anchor in which they are contained. 2894For example, all 2895.Ar anchor 2896rules specified in the main ruleset will reference anchor 2897attachment points underneath the main ruleset, and 2898.Ar anchor 2899rules specified in a file loaded from a 2900.Ar load anchor 2901rule will be attached under that anchor point. 2902.Pp 2903Rules may be contained in 2904.Ar anchor 2905attachment points which do not contain any rules when the main ruleset 2906is loaded, and later such anchors can be manipulated through 2907.Xr pfctl 8 2908without reloading the main ruleset or other anchors. 2909For example, 2910.Bd -literal -offset indent 2911ext_if = \&"kue0\&" 2912block on $ext_if all 2913anchor spam 2914pass out on $ext_if all 2915pass in on $ext_if proto tcp from any \e 2916 to $ext_if port smtp 2917.Ed 2918.Pp 2919blocks all packets on the external interface by default, then evaluates 2920all rules in the 2921.Ar anchor 2922named "spam", and finally passes all outgoing connections and 2923incoming connections to port 25. 2924.Bd -literal -offset indent 2925# echo \&"block in quick from 1.2.3.4 to any\&" \&| \e 2926 pfctl -a spam -f - 2927.Ed 2928.Pp 2929This loads a single rule into the 2930.Ar anchor , 2931which blocks all packets from a specific address. 2932.Pp 2933The anchor can also be populated by adding a 2934.Ar load anchor 2935rule after the 2936.Ar anchor 2937rule: 2938.Bd -literal -offset indent 2939anchor spam 2940load anchor spam from "/etc/pf-spam.conf" 2941.Ed 2942.Pp 2943When 2944.Xr pfctl 8 2945loads 2946.Nm pf.conf , 2947it will also load all the rules from the file 2948.Pa /etc/pf-spam.conf 2949into the anchor. 2950.Pp 2951Optionally, 2952.Ar anchor 2953rules can specify packet filtering parameters using the same syntax as 2954filter rules. 2955When parameters are used, the 2956.Ar anchor 2957rule is only evaluated for matching packets. 2958This allows conditional evaluation of anchors, like: 2959.Bd -literal -offset indent 2960block on $ext_if all 2961anchor spam proto tcp from any to any port smtp 2962pass out on $ext_if all 2963pass in on $ext_if proto tcp from any to $ext_if port smtp 2964.Ed 2965.Pp 2966The rules inside 2967.Ar anchor 2968spam are only evaluated for 2969.Ar tcp 2970packets with destination port 25. 2971Hence, 2972.Bd -literal -offset indent 2973# echo \&"block in quick from 1.2.3.4 to any" \&| \e 2974 pfctl -a spam -f - 2975.Ed 2976.Pp 2977will only block connections from 1.2.3.4 to port 25. 2978.Pp 2979Anchors may end with the asterisk 2980.Pq Sq * 2981character, which signifies that all anchors attached at that point 2982should be evaluated in the alphabetical ordering of their anchor name. 2983For example, 2984.Bd -literal -offset indent 2985anchor "spam/*" 2986.Ed 2987.Pp 2988will evaluate each rule in each anchor attached to the 2989.Li spam 2990anchor. 2991Note that it will only evaluate anchors that are directly attached to the 2992.Li spam 2993anchor, and will not descend to evaluate anchors recursively. 2994.Pp 2995Since anchors are evaluated relative to the anchor in which they are 2996contained, there is a mechanism for accessing the parent and ancestor 2997anchors of a given anchor. 2998Similar to file system path name resolution, if the sequence 2999.Dq .. 3000appears as an anchor path component, the parent anchor of the current 3001anchor in the path evaluation at that point will become the new current 3002anchor. 3003As an example, consider the following: 3004.Bd -literal -offset indent 3005# echo ' anchor "spam/allowed" ' | pfctl -f - 3006# echo -e ' anchor "../banned" \en pass' | \e 3007 pfctl -a spam/allowed -f - 3008.Ed 3009.Pp 3010Evaluation of the main ruleset will lead into the 3011.Li spam/allowed 3012anchor, which will evaluate the rules in the 3013.Li spam/banned 3014anchor, if any, before finally evaluating the 3015.Ar pass 3016rule. 3017.Pp 3018Filter rule 3019.Ar anchors 3020can also be loaded inline in the ruleset within a brace ('{' '}') delimited 3021block. 3022Brace delimited blocks may contain rules or other brace-delimited blocks. 3023When anchors are loaded this way the anchor name becomes optional. 3024.Bd -literal -offset indent 3025anchor "external" on $ext_if { 3026 block 3027 anchor out { 3028 pass proto tcp from any to port { 25, 80, 443 } 3029 } 3030 pass in proto tcp to any port 22 3031} 3032.Ed 3033.Pp 3034Since the parser specification for anchor names is a string, any 3035reference to an anchor name containing 3036.Sq / 3037characters will require double quote 3038.Pq Sq \&" 3039characters around the anchor name. 3040.Sh SCTP CONSIDERATIONS 3041.Xr pf 4 3042supports 3043.Xr sctp 4 3044connections. 3045It can match ports, track state and NAT SCTP traffic. 3046However, it will not alter port numbers during nat or rdr translations. 3047Doing so would break SCTP multihoming. 3048.Sh TRANSLATION EXAMPLES 3049This example maps incoming requests on port 80 to port 8080, on 3050which a daemon is running (because, for example, it is not run as root, 3051and therefore lacks permission to bind to port 80). 3052.Bd -literal 3053# use a macro for the interface name, so it can be changed easily 3054ext_if = \&"ne3\&" 3055 3056# map daemon on 8080 to appear to be on 80 3057rdr on $ext_if proto tcp from any to any port 80 -> 127.0.0.1 port 8080 3058.Ed 3059.Pp 3060If the 3061.Ar pass 3062modifier is given, packets matching the translation rule are passed without 3063inspecting the filter rules: 3064.Bd -literal 3065rdr pass on $ext_if proto tcp from any to any port 80 -> 127.0.0.1 \e 3066 port 8080 3067.Ed 3068.Pp 3069In the example below, vlan12 is configured as 192.168.168.1; 3070the machine translates all packets coming from 192.168.168.0/24 to 204.92.77.111 3071when they are going out any interface except vlan12. 3072This has the net effect of making traffic from the 192.168.168.0/24 3073network appear as though it is the Internet routable address 3074204.92.77.111 to nodes behind any interface on the router except 3075for the nodes on vlan12. 3076(Thus, 192.168.168.1 can talk to the 192.168.168.0/24 nodes.) 3077.Bd -literal 3078nat on ! vlan12 from 192.168.168.0/24 to any -> 204.92.77.111 3079.Ed 3080.Pp 3081In the example below, the machine sits between a fake internal 144.19.74.* 3082network, and a routable external IP of 204.92.77.100. 3083The 3084.Ar no nat 3085rule excludes protocol AH from being translated. 3086.Bd -literal 3087# NO NAT 3088no nat on $ext_if proto ah from 144.19.74.0/24 to any 3089nat on $ext_if from 144.19.74.0/24 to any -> 204.92.77.100 3090.Ed 3091.Pp 3092In the example below, packets bound for one specific server, as well as those 3093generated by the sysadmins are not proxied; all other connections are. 3094.Bd -literal 3095# NO RDR 3096no rdr on $int_if proto { tcp, udp } from any to $server port 80 3097no rdr on $int_if proto { tcp, udp } from $sysadmins to any port 80 3098rdr on $int_if proto { tcp, udp } from any to any port 80 -> 127.0.0.1 \e 3099 port 80 3100.Ed 3101.Pp 3102This longer example uses both a NAT and a redirection. 3103The external interface has the address 157.161.48.183. 3104On localhost, we are running 3105.Xr ftp-proxy 8 , 3106waiting for FTP sessions to be redirected to it. 3107The three mandatory anchors for 3108.Xr ftp-proxy 8 3109are omitted from this example; see the 3110.Xr ftp-proxy 8 3111manpage. 3112.Bd -literal 3113# NAT 3114# Translate outgoing packets' source addresses (any protocol). 3115# In this case, any address but the gateway's external address is mapped. 3116nat on $ext_if inet from ! ($ext_if) to any -> ($ext_if) 3117 3118# NAT PROXYING 3119# Map outgoing packets' source port to an assigned proxy port instead of 3120# an arbitrary port. 3121# In this case, proxy outgoing isakmp with port 500 on the gateway. 3122nat on $ext_if inet proto udp from any port = isakmp to any -> ($ext_if) \e 3123 port 500 3124 3125# BINAT 3126# Translate outgoing packets' source address (any protocol). 3127# Translate incoming packets' destination address to an internal machine 3128# (bidirectional). 3129binat on $ext_if from 10.1.2.150 to any -> $ext_if 3130 3131# Translate packets arriving on $peer_if addressed to 172.22.16.0/20 3132# to the corresponding address in 172.21.16.0/20 (bidirectional). 3133binat on $peer_if from 172.21.16.0/20 to any -> 172.22.16.0/20 3134 3135# RDR 3136# Translate incoming packets' destination addresses. 3137# As an example, redirect a TCP and UDP port to an internal machine. 3138rdr on $ext_if inet proto tcp from any to ($ext_if) port 8080 \e 3139 -> 10.1.2.151 port 22 3140rdr on $ext_if inet proto udp from any to ($ext_if) port 8080 \e 3141 -> 10.1.2.151 port 53 3142 3143# RDR 3144# Translate outgoing ftp control connections to send them to localhost 3145# for proxying with ftp-proxy(8) running on port 8021. 3146rdr on $int_if proto tcp from any to any port 21 -> 127.0.0.1 port 8021 3147.Ed 3148.Pp 3149In this example, a NAT gateway is set up to translate internal addresses 3150using a pool of public addresses (192.0.2.16/28) and to redirect 3151incoming web server connections to a group of web servers on the internal 3152network. 3153.Bd -literal 3154# NAT LOAD BALANCE 3155# Translate outgoing packets' source addresses using an address pool. 3156# A given source address is always translated to the same pool address by 3157# using the source-hash keyword. 3158nat on $ext_if inet from any to any -> 192.0.2.16/28 source-hash 3159 3160# RDR ROUND ROBIN 3161# Translate incoming web server connections to a group of web servers on 3162# the internal network. 3163rdr on $ext_if proto tcp from any to any port 80 \e 3164 -> { 10.1.2.155, 10.1.2.160, 10.1.2.161 } round-robin 3165.Ed 3166.Sh FILTER EXAMPLES 3167.Bd -literal 3168# The external interface is kue0 3169# (157.161.48.183, the only routable address) 3170# and the private network is 10.0.0.0/8, for which we are doing NAT. 3171 3172# Reassemble incoming traffic 3173set reassemble yes 3174 3175# use a macro for the interface name, so it can be changed easily 3176ext_if = \&"kue0\&" 3177 3178# block and log everything by default 3179block return log on $ext_if all 3180 3181# block anything coming from source we have no back routes for 3182block in from no-route to any 3183 3184# block packets whose ingress interface does not match the one in 3185# the route back to their source address 3186block in from urpf-failed to any 3187 3188# block and log outgoing packets that do not have our address as source, 3189# they are either spoofed or something is misconfigured (NAT disabled, 3190# for instance), we want to be nice and do not send out garbage. 3191block out log quick on $ext_if from ! 157.161.48.183 to any 3192 3193# silently drop broadcasts (cable modem noise) 3194block in quick on $ext_if from any to 255.255.255.255 3195 3196# block and log incoming packets from reserved address space and invalid 3197# addresses, they are either spoofed or misconfigured, we cannot reply to 3198# them anyway (hence, no return-rst). 3199block in log quick on $ext_if from { 10.0.0.0/8, 172.16.0.0/12, \e 3200 192.168.0.0/16, 255.255.255.255/32 } to any 3201 3202# ICMP 3203 3204# pass out/in certain ICMP queries and keep state (ping) 3205# state matching is done on host addresses and ICMP id (not type/code), 3206# so replies (like 0/0 for 8/0) will match queries 3207# ICMP error messages (which always refer to a TCP/UDP packet) are 3208# handled by the TCP/UDP states 3209pass on $ext_if inet proto icmp all icmp-type 8 code 0 3210 3211# UDP 3212 3213# pass out all UDP connections and keep state 3214pass out on $ext_if proto udp all 3215 3216# pass in certain UDP connections and keep state (DNS) 3217pass in on $ext_if proto udp from any to any port domain 3218 3219# TCP 3220 3221# pass out all TCP connections and modulate state 3222pass out on $ext_if proto tcp all modulate state 3223 3224# pass in certain TCP connections and keep state (SSH, SMTP, DNS, IDENT) 3225pass in on $ext_if proto tcp from any to any port { ssh, smtp, domain, \e 3226 auth } 3227 3228# Do not allow Windows 9x SMTP connections since they are typically 3229# a viral worm. Alternately we could limit these OSes to 1 connection each. 3230block in on $ext_if proto tcp from any os {"Windows 95", "Windows 98"} \e 3231 to any port smtp 3232 3233# IPv6 3234# pass in/out all IPv6 traffic: note that we have to enable this in two 3235# different ways, on both our physical interface and our tunnel 3236pass quick on gif0 inet6 3237pass quick on $ext_if proto ipv6 3238 3239# Packet Tagging 3240 3241# three interfaces: $int_if, $ext_if, and $wifi_if (wireless). NAT is 3242# being done on $ext_if for all outgoing packets. tag packets in on 3243# $int_if and pass those tagged packets out on $ext_if. all other 3244# outgoing packets (i.e., packets from the wireless network) are only 3245# permitted to access port 80. 3246 3247pass in on $int_if from any to any tag INTNET 3248pass in on $wifi_if from any to any 3249 3250block out on $ext_if from any to any 3251pass out quick on $ext_if tagged INTNET 3252pass out on $ext_if proto tcp from any to any port 80 3253 3254# tag incoming packets as they are redirected to spamd(8). use the tag 3255# to pass those packets through the packet filter. 3256 3257rdr on $ext_if inet proto tcp from <spammers> to port smtp \e 3258 tag SPAMD -> 127.0.0.1 port spamd 3259 3260block in on $ext_if 3261pass in on $ext_if inet proto tcp tagged SPAMD 3262.Ed 3263.Pp 3264In the example below, a router handling both address families 3265translates an internal IPv4 subnet to IPv6 using the well-known 326664:ff9b::/96 prefix: 3267.Bd -literal -offset 4n 3268pass in on $v4_if inet af-to inet6 from ($v6_if) to 64:ff9b::/96 3269.Ed 3270.Pp 3271Paired with the example above, the example below can be used on 3272another router handling both address families to translate back 3273to IPv4: 3274.Bd -literal -offset 4n 3275pass in on $v6_if inet6 to 64:ff9b::/96 af-to inet from ($v4_if) 3276.Ed 3277.Sh GRAMMAR 3278Syntax for 3279.Nm 3280in BNF: 3281.Bd -literal 3282line = ( option | ether-rule | pf-rule | nat-rule | binat-rule | 3283 rdr-rule | antispoof-rule | altq-rule | queue-rule | 3284 trans-anchors | anchor-rule | anchor-close | load-anchor | 3285 table-rule | include ) 3286 3287option = "set" ( [ "timeout" ( timeout | "{" timeout-list "}" ) ] | 3288 [ "ruleset-optimization" [ "none" | "basic" | "profile" ]] | 3289 [ "optimization" [ "default" | "normal" | 3290 "high-latency" | "satellite" | 3291 "aggressive" | "conservative" ] ] 3292 [ "limit" ( limit-item | "{" limit-list "}" ) ] | 3293 [ "loginterface" ( interface-name | "none" ) ] | 3294 [ "block-policy" ( "drop" | "return" ) ] | 3295 [ "state-policy" ( "if-bound" | "floating" ) ] 3296 [ "state-defaults" state-opts ] 3297 [ "require-order" ( "yes" | "no" ) ] 3298 [ "fingerprints" filename ] | 3299 [ "skip on" ifspec ] | 3300 [ "debug" ( "none" | "urgent" | "misc" | "loud" ) ] 3301 [ "keepcounters" ] ) 3302 3303ether-rule = "ether" etheraction [ ( "in" | "out" ) ] 3304 [ "quick" ] [ "on" ifspec ] [ "bridge-to" interface-name ] 3305 [ etherprotospec ] etherhosts [ "l3" hosts ] 3306 [ etherfilteropt-list ] 3307 3308pf-rule = action [ ( "in" | "out" ) ] 3309 [ "log" [ "(" logopts ")"] ] [ "quick" ] 3310 [ "on" ifspec ] [ route ] [ af ] [ protospec ] 3311 hosts [ filteropt-list ] 3312 3313logopts = logopt [ "," logopts ] 3314logopt = "all" | "matches" | "user" | "to" interface-name 3315 3316etherfilteropt-list = etherfilteropt-list etherfilteropt | etherfilteropt 3317etherfilteropt = "tag" string | "tagged" string | "queue" ( string ) | 3318 "ridentifier" number | "label" string 3319 3320filteropt-list = filteropt-list filteropt | filteropt 3321filteropt = user | group | flags | icmp-type | icmp6-type | "tos" tos | 3322 "af-to" af "from" ( redirhost | "{" redirhost-list "}" ) 3323 [ "to" ( redirhost | "{" redirhost-list "}" ) ] | 3324 ( "no" | "keep" | "modulate" | "synproxy" ) "state" 3325 [ "(" state-opts ")" ] | 3326 "fragment" | "no-df" | "min-ttl" number | "set-tos" tos | 3327 "max-mss" number | "random-id" | "reassemble tcp" | 3328 fragmentation | "allow-opts" | 3329 "label" string | "tag" string | [ ! ] "tagged" string | 3330 "set prio" ( number | "(" number [ [ "," ] number ] ")" ) | 3331 "queue" ( string | "(" string [ [ "," ] string ] ")" ) | 3332 "rtable" number | "probability" number"%" | "prio" number | 3333 "dnpipe" ( number | "(" number "," number ")" ) | 3334 "dnqueue" ( number | "(" number "," number ")" ) | 3335 "ridentifier" number | 3336 [ ! ] "received-on" ( interface-name | interface-group ) 3337 3338nat-rule = [ "no" ] "nat" [ "pass" [ "log" [ "(" logopts ")" ] ] ] 3339 [ "on" ifspec ] [ af ] 3340 [ protospec ] hosts [ "tag" string ] [ "tagged" string ] 3341 [ "->" ( redirhost | "{" redirhost-list "}" ) 3342 [ portspec ] [ pooltype ] [ "static-port" ] 3343 [ "map-e-portset" number "/" number "/" number ] ] 3344 3345binat-rule = [ "no" ] "binat" [ "pass" [ "log" [ "(" logopts ")" ] ] ] 3346 [ "on" interface-name ] [ af ] 3347 [ "proto" ( proto-name | proto-number ) ] 3348 "from" address [ "/" mask-bits ] "to" ipspec 3349 [ "tag" string ] [ "tagged" string ] 3350 [ "->" address [ "/" mask-bits ] ] 3351 3352rdr-rule = [ "no" ] "rdr" [ "pass" [ "log" [ "(" logopts ")" ] ] ] 3353 [ "on" ifspec ] [ af ] 3354 [ protospec ] hosts [ "tag" string ] [ "tagged" string ] 3355 [ "->" ( redirhost | "{" redirhost-list "}" ) 3356 [ portspec ] [ pooltype ] ] 3357 3358antispoof-rule = "antispoof" [ "log" ] [ "quick" ] 3359 "for" ifspec [ af ] [ "label" string ] 3360 [ "ridentifier" number ] 3361 3362table-rule = "table" "<" string ">" [ tableopts-list ] 3363tableopts-list = tableopts-list tableopts | tableopts 3364tableopts = "persist" | "const" | "counters" | "file" string | 3365 "{" [ tableaddr-list ] "}" 3366tableaddr-list = tableaddr-list [ "," ] tableaddr-spec | tableaddr-spec 3367tableaddr-spec = [ "!" ] tableaddr [ "/" mask-bits ] 3368tableaddr = hostname | ifspec | "self" | 3369 ipv4-dotted-quad | ipv6-coloned-hex 3370 3371altq-rule = "altq on" interface-name queueopts-list 3372 "queue" subqueue 3373queue-rule = "queue" string [ "on" interface-name ] queueopts-list 3374 subqueue 3375 3376anchor-rule = "anchor" [ string ] [ ( "in" | "out" ) ] [ "on" ifspec ] 3377 [ af ] [ protospec ] [ hosts ] [ filteropt-list ] [ "{" ] 3378 3379anchor-close = "}" 3380 3381trans-anchors = ( "nat-anchor" | "rdr-anchor" | "binat-anchor" ) string 3382 [ "on" ifspec ] [ af ] [ "proto" ] [ protospec ] [ hosts ] 3383 3384load-anchor = "load anchor" string "from" filename 3385 3386queueopts-list = queueopts-list queueopts | queueopts 3387queueopts = [ "bandwidth" bandwidth-spec ] | 3388 [ "qlimit" number ] | [ "tbrsize" number ] | 3389 [ "priority" number ] | [ schedulers ] 3390schedulers = ( cbq-def | priq-def | hfsc-def ) 3391bandwidth-spec = "number" ( "b" | "Kb" | "Mb" | "Gb" | "%" ) 3392 3393etheraction = "pass" | "block" 3394action = "pass" | "match" | "block" [ return ] | [ "no" ] "scrub" 3395return = "drop" | "return" | "return-rst" [ "( ttl" number ")" ] | 3396 "return-icmp" [ "(" icmpcode [ [ "," ] icmp6code ] ")" ] | 3397 "return-icmp6" [ "(" icmp6code ")" ] 3398icmpcode = ( icmp-code-name | icmp-code-number ) 3399icmp6code = ( icmp6-code-name | icmp6-code-number ) 3400 3401ifspec = ( [ "!" ] ( interface-name | interface-group ) ) | 3402 "{" interface-list "}" 3403interface-list = [ "!" ] ( interface-name | interface-group ) 3404 [ [ "," ] interface-list ] 3405route = ( "route-to" | "reply-to" | "dup-to" ) 3406 ( routehost | "{" routehost-list "}" ) 3407 [ pooltype ] 3408af = "inet" | "inet6" 3409 3410etherprotospec = "proto" ( proto-number | "{" etherproto-list "}" ) 3411etherproto-list = proto-number [ [ "," ] etherproto-list ] 3412protospec = "proto" ( proto-name | proto-number | 3413 "{" proto-list "}" ) 3414proto-list = ( proto-name | proto-number ) [ [ "," ] proto-list ] 3415 3416etherhosts = "from" macaddress "to" macaddress 3417macaddress = mac | mac "/" masklen | mac "&" mask 3418 3419hosts = "all" | 3420 "from" ( "any" | "no-route" | "urpf-failed" | "self" | host | 3421 "{" host-list "}" ) [ port ] [ os ] 3422 "to" ( "any" | "no-route" | "self" | host | 3423 "{" host-list "}" ) [ port ] 3424 3425ipspec = "any" | host | "{" host-list "}" 3426host = [ "!" ] ( address [ "/" mask-bits ] | "<" string ">" ) 3427redirhost = address [ "/" mask-bits ] 3428routehost = "(" interface-name [ address [ "/" mask-bits ] ] ")" 3429address = ( interface-name | interface-group | 3430 "(" ( interface-name | interface-group ) ")" | 3431 hostname | ipv4-dotted-quad | ipv6-coloned-hex ) 3432host-list = host [ [ "," ] host-list ] 3433redirhost-list = redirhost [ [ "," ] redirhost-list ] 3434routehost-list = routehost [ [ "," ] routehost-list ] 3435 3436port = "port" ( unary-op | binary-op | "{" op-list "}" ) 3437portspec = "port" ( number | name ) [ ":" ( "*" | number | name ) ] 3438os = "os" ( os-name | "{" os-list "}" ) 3439user = "user" ( unary-op | binary-op | "{" op-list "}" ) 3440group = "group" ( unary-op | binary-op | "{" op-list "}" ) 3441 3442unary-op = [ "=" | "!=" | "<" | "<=" | ">" | ">=" ] 3443 ( name | number ) 3444binary-op = number ( "<>" | "><" | ":" ) number 3445op-list = ( unary-op | binary-op ) [ [ "," ] op-list ] 3446 3447os-name = operating-system-name 3448os-list = os-name [ [ "," ] os-list ] 3449 3450flags = "flags" ( [ flag-set ] "/" flag-set | "any" ) 3451flag-set = [ "F" ] [ "S" ] [ "R" ] [ "P" ] [ "A" ] [ "U" ] [ "E" ] 3452 [ "W" ] 3453 3454icmp-type = "icmp-type" ( icmp-type-code | "{" icmp-list "}" ) 3455icmp6-type = "icmp6-type" ( icmp-type-code | "{" icmp-list "}" ) 3456icmp-type-code = ( icmp-type-name | icmp-type-number ) 3457 [ "code" ( icmp-code-name | icmp-code-number ) ] 3458icmp-list = icmp-type-code [ [ "," ] icmp-list ] 3459 3460tos = ( "lowdelay" | "throughput" | "reliability" | 3461 [ "0x" ] number ) 3462 3463state-opts = state-opt [ [ "," ] state-opts ] 3464state-opt = ( "max" number | "no-sync" | timeout | "sloppy" | 3465 "source-track" [ ( "rule" | "global" ) ] | 3466 "max-src-nodes" number | "max-src-states" number | 3467 "max-src-conn" number | 3468 "max-src-conn-rate" number "/" number | 3469 "overload" "<" string ">" [ "flush" ] | 3470 "if-bound" | "floating" | "pflow" ) 3471 3472fragmentation = [ "fragment reassemble" ] 3473 3474timeout-list = timeout [ [ "," ] timeout-list ] 3475timeout = ( "tcp.first" | "tcp.opening" | "tcp.established" | 3476 "tcp.closing" | "tcp.finwait" | "tcp.closed" | 3477 "sctp.first" | "sctp.opening" | "sctp.established" | 3478 "sctp.closing" | "sctp.closed" | 3479 "udp.first" | "udp.single" | "udp.multiple" | 3480 "icmp.first" | "icmp.error" | 3481 "other.first" | "other.single" | "other.multiple" | 3482 "frag" | "interval" | "src.track" | 3483 "adaptive.start" | "adaptive.end" ) number 3484 3485limit-list = limit-item [ [ "," ] limit-list ] 3486limit-item = ( "states" | "frags" | "src-nodes" ) number 3487 3488pooltype = ( "bitmask" | "random" | 3489 "source-hash" [ ( hex-key | string-key ) ] | 3490 "round-robin" ) [ sticky-address ] 3491 3492subqueue = string | "{" queue-list "}" 3493queue-list = string [ [ "," ] string ] 3494cbq-def = "cbq" [ "(" cbq-opt [ [ "," ] cbq-opt ] ")" ] 3495priq-def = "priq" [ "(" priq-opt [ [ "," ] priq-opt ] ")" ] 3496hfsc-def = "hfsc" [ "(" hfsc-opt [ [ "," ] hfsc-opt ] ")" ] 3497cbq-opt = ( "default" | "borrow" | "red" | "ecn" | "rio" ) 3498priq-opt = ( "default" | "red" | "ecn" | "rio" ) 3499hfsc-opt = ( "default" | "red" | "ecn" | "rio" | 3500 linkshare-sc | realtime-sc | upperlimit-sc ) 3501linkshare-sc = "linkshare" sc-spec 3502realtime-sc = "realtime" sc-spec 3503upperlimit-sc = "upperlimit" sc-spec 3504sc-spec = ( bandwidth-spec | 3505 "(" bandwidth-spec number bandwidth-spec ")" ) 3506include = "include" filename 3507.Ed 3508.Sh FILES 3509.Bl -tag -width "/etc/protocols" -compact 3510.It Pa /etc/hosts 3511Host name database. 3512.It Pa /etc/pf.conf 3513Default location of the ruleset file. 3514The file has to be created manually as it is not installed with a 3515standard installation. 3516.It Pa /etc/pf.os 3517Default location of OS fingerprints. 3518.It Pa /etc/protocols 3519Protocol name database. 3520.It Pa /etc/services 3521Service name database. 3522.El 3523.Sh SEE ALSO 3524.Xr altq 4 , 3525.Xr carp 4 , 3526.Xr icmp 4 , 3527.Xr icmp6 4 , 3528.Xr ip 4 , 3529.Xr ip6 4 , 3530.Xr pf 4 , 3531.Xr pflow 4 , 3532.Xr pfsync 4 , 3533.Xr sctp 4 , 3534.Xr tcp 4 , 3535.Xr udp 4 , 3536.Xr hosts 5 , 3537.Xr pf.os 5 , 3538.Xr protocols 5 , 3539.Xr services 5 , 3540.Xr ftp-proxy 8 , 3541.Xr pfctl 8 , 3542.Xr pflogd 8 3543.Sh HISTORY 3544The 3545.Nm 3546file format first appeared in 3547.Ox 3.0 . 3548